Blame - docs/ProgrammersManual.html - fp2-dev/platform/external/llvm

blob: 036c387d7a592cfca399803199f78907ab44944a [file] [log] [blame]

Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	1	<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
				2	"http://www.w3.org/TR/html4/strict.dtd">
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3	<html>
				4	<head>
Owen Anderson	5e8c50e	2009-06-16 17:40:28 +0000	[diff] [blame]	5	<meta http-equiv="Content-type" content="text/html;charset=UTF-8">
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	6	<title>LLVM Programmer's Manual</title>
Daniel Dunbar	adea497	2012-04-19 20:20:34 +0000	[diff] [blame]	7	<link rel="stylesheet" href="_static/llvm.css" type="text/css">
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	8	</head>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	9	<body>
				10
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	11	<h1>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	12	LLVM Programmer's Manual
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	13	</h1>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	14
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	15	<ol>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	16	<li><a href="#introduction">Introduction</a></li>
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	17	<li><a href="#general">General Information</a>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	18	<ul>
Reid Spencer	fe8f4ff	2004-11-01 09:02:53 +0000	[diff] [blame]	19	<li><a href="#stl">The C++ Standard Template Library</a></li>
				20	<!--
				21	<li>The <tt>-time-passes</tt> option</li>
				22	<li>How to use the LLVM Makefile system</li>
				23	<li>How to write a regression test</li>
Chris Lattner	61db465	2004-12-08 19:05:44 +0000	[diff] [blame]	24
Reid Spencer	fe8f4ff	2004-11-01 09:02:53 +0000	[diff] [blame]	25	-->
Chris Lattner	84b7f8d	2003-08-01 22:20:59 +0000	[diff] [blame]	26	</ul>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	27	</li>
				28	<li><a href="#apis">Important and useful LLVM APIs</a>
				29	<ul>
				30	<li><a href="#isa">The <tt>isa<></tt>, <tt>cast<></tt>
				31	and <tt>dyn_cast<></tt> templates</a> </li>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	32	<li><a href="#string_apis">Passing strings (the <tt>StringRef</tt>
Benjamin Kramer	e15192b	2009-08-05 15:42:44 +0000	[diff] [blame]	33	and <tt>Twine</tt> classes)</a>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	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>
Benjamin Kramer	e15192b	2009-08-05 15:42:44 +0000	[diff] [blame]	38	</li>
Misha Brukman	2c122ce	2005-11-01 21:12:49 +0000	[diff] [blame]	39	<li><a href="#DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	40	option</a>
				41	<ul>
				42	<li><a href="#DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt>
				43	and the <tt>-debug-only</tt> option</a> </li>
				44	</ul>
				45	</li>
Chris Lattner	0be6fdf	2006-12-19 21:46:21 +0000	[diff] [blame]	46	<li><a href="#Statistic">The <tt>Statistic</tt> class & <tt>-stats</tt>
Reid Spencer	fe8f4ff	2004-11-01 09:02:53 +0000	[diff] [blame]	47	option</a></li>
				48	<!--
				49	<li>The <tt>InstVisitor</tt> template
				50	<li>The general graph API
				51	-->
Chris Lattner	f623a08	2005-10-17 01:36:23 +0000	[diff] [blame]	52	<li><a href="#ViewGraph">Viewing graphs while debugging code</a></li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	53	</ul>
				54	</li>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	55	<li><a href="#datastructure">Picking the Right Data Structure for a Task</a>
				56	<ul>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	57	<li><a href="#ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
				58	<ul>
Chris Lattner	8ae4261	2011-04-05 23:18:20 +0000	[diff] [blame]	59	<li><a href="#dss_arrayref">llvm/ADT/ArrayRef.h</a></li>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	60	<li><a href="#dss_fixedarrays">Fixed Size Arrays</a></li>
				61	<li><a href="#dss_heaparrays">Heap Allocated Arrays</a></li>
Chris Lattner	9d69d4a	2011-07-18 01:40:02 +0000	[diff] [blame]	62	<li><a href="#dss_tinyptrvector">"llvm/ADT/TinyPtrVector.h"</a></li>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	63	<li><a href="#dss_smallvector">"llvm/ADT/SmallVector.h"</a></li>
				64	<li><a href="#dss_vector"><vector></a></li>
				65	<li><a href="#dss_deque"><deque></a></li>
				66	<li><a href="#dss_list"><list></a></li>
Gabor Greif	3899e49	2009-02-27 11:37:41 +0000	[diff] [blame]	67	<li><a href="#dss_ilist">llvm/ADT/ilist.h</a></li>
Argyrios Kyrtzidis	81df089	2011-06-15 19:56:01 +0000	[diff] [blame]	68	<li><a href="#dss_packedvector">llvm/ADT/PackedVector.h</a></li>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	69	<li><a href="#dss_other">Other Sequential Container Options</a></li>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	70	</ul></li>
Chris Lattner	7314a20	2011-07-22 20:46:49 +0000	[diff] [blame]	71	<li><a href="#ds_string">String-like containers</a>
Chris Lattner	6682746	2011-07-22 21:36:29 +0000	[diff] [blame]	72	<ul>
				73	<li><a href="#dss_stringref">llvm/ADT/StringRef.h</a></li>
				74	<li><a href="#dss_twine">llvm/ADT/Twine.h</a></li>
				75	<li><a href="#dss_smallstring">llvm/ADT/SmallString.h</a></li>
				76	<li><a href="#dss_stdstring">std::string</a></li>
				77	</ul></li>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	78	<li><a href="#ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
				79	<ul>
				80	<li><a href="#dss_sortedvectorset">A sorted 'vector'</a></li>
				81	<li><a href="#dss_smallset">"llvm/ADT/SmallSet.h"</a></li>
				82	<li><a href="#dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a></li>
Chris Lattner	c28476f	2007-09-30 00:58:59 +0000	[diff] [blame]	83	<li><a href="#dss_denseset">"llvm/ADT/DenseSet.h"</a></li>
Jakob Stoklund Olesen	6258862	2012-02-22 00:56:08 +0000	[diff] [blame]	84	<li><a href="#dss_sparseset">"llvm/ADT/SparseSet.h"</a></li>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	85	<li><a href="#dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a></li>
				86	<li><a href="#dss_set"><set></a></li>
				87	<li><a href="#dss_setvector">"llvm/ADT/SetVector.h"</a></li>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	88	<li><a href="#dss_uniquevector">"llvm/ADT/UniqueVector.h"</a></li>
Chris Lattner	2fdd005	2011-11-15 22:40:14 +0000	[diff] [blame]	89	<li><a href="#dss_immutableset">"llvm/ADT/ImmutableSet.h"</a></li>
				90	<li><a href="#dss_otherset">Other Set-Like Container Options</a></li>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	91	</ul></li>
Chris Lattner	f369252	2007-02-03 19:51:56 +0000	[diff] [blame]	92	<li><a href="#ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
				93	<ul>
				94	<li><a href="#dss_sortedvectormap">A sorted 'vector'</a></li>
Chris Lattner	796f9fa	2007-02-08 19:14:21 +0000	[diff] [blame]	95	<li><a href="#dss_stringmap">"llvm/ADT/StringMap.h"</a></li>
Chris Lattner	f369252	2007-02-03 19:51:56 +0000	[diff] [blame]	96	<li><a href="#dss_indexedmap">"llvm/ADT/IndexedMap.h"</a></li>
				97	<li><a href="#dss_densemap">"llvm/ADT/DenseMap.h"</a></li>
Jeffrey Yasskin	71a5c22	2009-10-22 22:11:22 +0000	[diff] [blame]	98	<li><a href="#dss_valuemap">"llvm/ADT/ValueMap.h"</a></li>
Jakob Stoklund Olesen	aca0da6	2010-12-14 00:55:51 +0000	[diff] [blame]	99	<li><a href="#dss_intervalmap">"llvm/ADT/IntervalMap.h"</a></li>
Chris Lattner	f369252	2007-02-03 19:51:56 +0000	[diff] [blame]	100	<li><a href="#dss_map"><map></a></li>
Jakob Stoklund Olesen	4359e5e	2011-04-05 20:56:08 +0000	[diff] [blame]	101	<li><a href="#dss_inteqclasses">"llvm/ADT/IntEqClasses.h"</a></li>
Chris Lattner	2fdd005	2011-11-15 22:40:14 +0000	[diff] [blame]	102	<li><a href="#dss_immutablemap">"llvm/ADT/ImmutableMap.h"</a></li>
Chris Lattner	f369252	2007-02-03 19:51:56 +0000	[diff] [blame]	103	<li><a href="#dss_othermap">Other Map-Like Container Options</a></li>
				104	</ul></li>
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	105	<li><a href="#ds_bit">BitVector-like containers</a>
				106	<ul>
				107	<li><a href="#dss_bitvector">A dense bitvector</a></li>
Dan Gohman	5f7775c	2010-01-05 18:24:00 +0000	[diff] [blame]	108	<li><a href="#dss_smallbitvector">A "small" dense bitvector</a></li>
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	109	<li><a href="#dss_sparsebitvector">A sparse bitvector</a></li>
				110	</ul></li>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	111	</ul>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	112	</li>
Chris Lattner	ae7f759	2002-09-06 18:31:18 +0000	[diff] [blame]	113	<li><a href="#common">Helpful Hints for Common Operations</a>
Chris Lattner	ae7f759	2002-09-06 18:31:18 +0000	[diff] [blame]	114	<ul>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	115	<li><a href="#inspection">Basic Inspection and Traversal Routines</a>
				116	<ul>
				117	<li><a href="#iterate_function">Iterating over the <tt>BasicBlock</tt>s
				118	in a <tt>Function</tt></a> </li>
				119	<li><a href="#iterate_basicblock">Iterating over the <tt>Instruction</tt>s
				120	in a <tt>BasicBlock</tt></a> </li>
				121	<li><a href="#iterate_institer">Iterating over the <tt>Instruction</tt>s
				122	in a <tt>Function</tt></a> </li>
				123	<li><a href="#iterate_convert">Turning an iterator into a
				124	class pointer</a> </li>
				125	<li><a href="#iterate_complex">Finding call sites: a more
				126	complex example</a> </li>
				127	<li><a href="#calls_and_invokes">Treating calls and invokes
				128	the same way</a> </li>
				129	<li><a href="#iterate_chains">Iterating over def-use &
				130	use-def chains</a> </li>
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	131	<li><a href="#iterate_preds">Iterating over predecessors &
				132	successors of blocks</a></li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	133	</ul>
				134	</li>
				135	<li><a href="#simplechanges">Making simple changes</a>
				136	<ul>
				137	<li><a href="#schanges_creating">Creating and inserting new
				138	<tt>Instruction</tt>s</a> </li>
				139	<li><a href="#schanges_deleting">Deleting <tt>Instruction</tt>s</a> </li>
				140	<li><a href="#schanges_replacing">Replacing an <tt>Instruction</tt>
				141	with another <tt>Value</tt></a> </li>
Tanya Lattner	b011c66	2007-06-20 18:33:15 +0000	[diff] [blame]	142	<li><a href="#schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a> </li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	143	</ul>
Reid Spencer	fe8f4ff	2004-11-01 09:02:53 +0000	[diff] [blame]	144	</li>
Jeffrey Yasskin	714257f	2009-04-30 22:33:41 +0000	[diff] [blame]	145	<li><a href="#create_types">How to Create Types</a></li>
Chris Lattner	ae7f759	2002-09-06 18:31:18 +0000	[diff] [blame]	146	<!--
				147	<li>Working with the Control Flow Graph
				148	<ul>
				149	<li>Accessing predecessors and successors of a <tt>BasicBlock</tt>
				150	<li>
				151	<li>
				152	</ul>
Reid Spencer	fe8f4ff	2004-11-01 09:02:53 +0000	[diff] [blame]	153	-->
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	154	</ul>
				155	</li>
Chris Lattner	d9d6e10	2005-04-23 16:10:52 +0000	[diff] [blame]	156
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	157	<li><a href="#threading">Threads and LLVM</a>
				158	<ul>
Owen Anderson	1ad70e3	2009-06-16 18:04:19 +0000	[diff] [blame]	159	<li><a href="#startmultithreaded">Entering and Exiting Multithreaded Mode
				160	</a></li>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	161	<li><a href="#shutdown">Ending execution with <tt>llvm_shutdown()</tt></a></li>
				162	<li><a href="#managedstatic">Lazy initialization with <tt>ManagedStatic</tt></a></li>
Owen Anderson	e0c951a	2009-08-19 17:58:52 +0000	[diff] [blame]	163	<li><a href="#llvmcontext">Achieving Isolation with <tt>LLVMContext</tt></a></li>
Jeffrey Yasskin	01eba39	2010-01-29 19:10:38 +0000	[diff] [blame]	164	<li><a href="#jitthreading">Threads and the JIT</a></li>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	165	</ul>
				166	</li>
				167
Chris Lattner	d9d6e10	2005-04-23 16:10:52 +0000	[diff] [blame]	168	<li><a href="#advanced">Advanced Topics</a>
				169	<ul>
Chris Lattner	f1b200b	2005-04-23 17:27:36 +0000	[diff] [blame]	170
Chris Lattner	1afcace	2011-07-09 17:41:24 +0000	[diff] [blame]	171	<li><a href="#SymbolTable">The <tt>ValueSymbolTable</tt> class</a></li>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	172	<li><a href="#UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a></li>
Chris Lattner	d9d6e10	2005-04-23 16:10:52 +0000	[diff] [blame]	173	</ul></li>
				174
Joel Stanley	9b96c44	2002-09-06 21:55:13 +0000	[diff] [blame]	175	<li><a href="#coreclasses">The Core LLVM Class Hierarchy Reference</a>
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	176	<ul>
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	177	<li><a href="#Type">The <tt>Type</tt> class</a> </li>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	178	<li><a href="#Module">The <tt>Module</tt> class</a></li>
Reid Spencer	fe8f4ff	2004-11-01 09:02:53 +0000	[diff] [blame]	179	<li><a href="#Value">The <tt>Value</tt> class</a>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	180	<ul>
				181	<li><a href="#User">The <tt>User</tt> class</a>
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	182	<ul>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	183	<li><a href="#Instruction">The <tt>Instruction</tt> class</a></li>
				184	<li><a href="#Constant">The <tt>Constant</tt> class</a>
				185	<ul>
				186	<li><a href="#GlobalValue">The <tt>GlobalValue</tt> class</a>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	187	<ul>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	188	<li><a href="#Function">The <tt>Function</tt> class</a></li>
				189	<li><a href="#GlobalVariable">The <tt>GlobalVariable</tt> class</a></li>
				190	</ul>
				191	</li>
				192	</ul>
				193	</li>
Reid Spencer	fe8f4ff	2004-11-01 09:02:53 +0000	[diff] [blame]	194	</ul>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	195	</li>
				196	<li><a href="#BasicBlock">The <tt>BasicBlock</tt> class</a></li>
				197	<li><a href="#Argument">The <tt>Argument</tt> class</a></li>
				198	</ul>
Reid Spencer	fe8f4ff	2004-11-01 09:02:53 +0000	[diff] [blame]	199	</li>
				200	</ul>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	201	</li>
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	202	</ol>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	203
Chris Lattner	69bf8a9	2004-05-23 21:06:58 +0000	[diff] [blame]	204	<div class="doc_author">
				205	<p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>,
Chris Lattner	94c4359	2004-05-26 16:52:55 +0000	[diff] [blame]	206	<a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a>,
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	207	<a href="mailto:ggreif@gmail.com">Gabor Greif</a>,
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	208	<a href="mailto:jstanley@cs.uiuc.edu">Joel Stanley</a>,
				209	<a href="mailto:rspencer@x10sys.com">Reid Spencer</a> and
				210	<a href="mailto:owen@apple.com">Owen Anderson</a></p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	211	</div>
				212
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	213	<!-- *********************************************************************** -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	214	<h2>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	215	<a name="introduction">Introduction </a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	216	</h2>
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	217	<!-- *********************************************************************** -->
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	218
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	219	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	220
				221	<p>This document is meant to highlight some of the important classes and
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	222	interfaces available in the LLVM source-base. This manual is not
				223	intended to explain what LLVM is, how it works, and what LLVM code looks
				224	like. It assumes that you know the basics of LLVM and are interested
				225	in writing transformations or otherwise analyzing or manipulating the
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	226	code.</p>
				227
				228	<p>This document should get you oriented so that you can find your
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	229	way in the continuously growing source code that makes up the LLVM
				230	infrastructure. Note that this manual is not intended to serve as a
				231	replacement for reading the source code, so if you think there should be
				232	a method in one of these classes to do something, but it's not listed,
				233	check the source. Links to the <a href="/doxygen/">doxygen</a> sources
				234	are provided to make this as easy as possible.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	235
				236	<p>The first section of this document describes general information that is
				237	useful to know when working in the LLVM infrastructure, and the second describes
				238	the Core LLVM classes. In the future this manual will be extended with
				239	information describing how to use extension libraries, such as dominator
				240	information, CFG traversal routines, and useful utilities like the <tt><a
				241	href="/doxygen/InstVisitor_8h-source.html">InstVisitor</a></tt> template.</p>
				242
				243	</div>
				244
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	245	<!-- *********************************************************************** -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	246	<h2>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	247	<a name="general">General Information</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	248	</h2>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	249	<!-- *********************************************************************** -->
				250
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	251	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	252
				253	<p>This section contains general information that is useful if you are working
				254	in the LLVM source-base, but that isn't specific to any particular API.</p>
				255
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	256	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	257	<h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	258	<a name="stl">The C++ Standard Template Library</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	259	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	260
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	261	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	262
				263	<p>LLVM makes heavy use of the C++ Standard Template Library (STL),
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	264	perhaps much more than you are used to, or have seen before. Because of
				265	this, you might want to do a little background reading in the
				266	techniques used and capabilities of the library. There are many good
				267	pages that discuss the STL, and several books on the subject that you
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	268	can get, so it will not be discussed in this document.</p>
				269
				270	<p>Here are some useful links:</p>
				271
				272	<ol>
				273
Nick Lewycky	ea1fe2c	2010-10-09 21:12:29 +0000	[diff] [blame]	274	<li><a href="http://www.dinkumware.com/manuals/#Standard C++ Library">Dinkumware
				275	C++ Library reference</a> - an excellent reference for the STL and other parts
				276	of the standard C++ library.</li>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	277
				278	<li><a href="http://www.tempest-sw.com/cpp/">C++ In a Nutshell</a> - This is an
Gabor Greif	0cbcabe	2009-03-12 09:47:03 +0000	[diff] [blame]	279	O'Reilly book in the making. It has a decent Standard Library
				280	Reference that rivals Dinkumware's, and is unfortunately no longer free since the
				281	book has been published.</li>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	282
				283	<li><a href="http://www.parashift.com/c++-faq-lite/">C++ Frequently Asked
				284	Questions</a></li>
				285
				286	<li><a href="http://www.sgi.com/tech/stl/">SGI's STL Programmer's Guide</a> -
				287	Contains a useful <a
				288	href="http://www.sgi.com/tech/stl/stl_introduction.html">Introduction to the
				289	STL</a>.</li>
				290
				291	<li><a href="http://www.research.att.com/%7Ebs/C++.html">Bjarne Stroustrup's C++
				292	Page</a></li>
				293
Tanya Lattner	79445ba	2004-12-08 18:34:56 +0000	[diff] [blame]	294	<li><a href="http://64.78.49.204/">
Reid Spencer	096603a	2004-05-26 08:41:35 +0000	[diff] [blame]	295	Bruce Eckel's Thinking in C++, 2nd ed. Volume 2 Revision 4.0 (even better, get
				296	the book).</a></li>
				297
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	298	</ol>
				299
				300	<p>You are also encouraged to take a look at the <a
				301	href="CodingStandards.html">LLVM Coding Standards</a> guide which focuses on how
				302	to write maintainable code more than where to put your curly braces.</p>
				303
				304	</div>
				305
				306	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	307	<h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	308	<a name="stl">Other useful references</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	309	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	310
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	311	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	312
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	313	<ol>
Misha Brukman	a0f71e4	2004-06-18 18:39:00 +0000	[diff] [blame]	314	<li><a href="http://www.fortran-2000.com/ArnaudRecipes/sharedlib.html">Using
				315	static and shared libraries across platforms</a></li>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	316	</ol>
				317
				318	</div>
				319
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	320	</div>
				321
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	322	<!-- *********************************************************************** -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	323	<h2>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	324	<a name="apis">Important and useful LLVM APIs</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	325	</h2>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	326	<!-- *********************************************************************** -->
				327
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	328	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	329
				330	<p>Here we highlight some LLVM APIs that are generally useful and good to
				331	know about when writing transformations.</p>
				332
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	333	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	334	<h3>
Misha Brukman	2c122ce	2005-11-01 21:12:49 +0000	[diff] [blame]	335	<a name="isa">The <tt>isa<></tt>, <tt>cast<></tt> and
				336	<tt>dyn_cast<></tt> templates</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	337	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	338
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	339	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	340
				341	<p>The LLVM source-base makes extensive use of a custom form of RTTI.
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	342	These templates have many similarities to the C++ <tt>dynamic_cast<></tt>
				343	operator, but they don't have some drawbacks (primarily stemming from
				344	the fact that <tt>dynamic_cast<></tt> only works on classes that
				345	have a v-table). Because they are used so often, you must know what they
				346	do and how they work. All of these templates are defined in the <a
Chris Lattner	695b78b	2005-04-26 22:56:16 +0000	[diff] [blame]	347	href="/doxygen/Casting_8h-source.html"><tt>llvm/Support/Casting.h</tt></a>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	348	file (note that you very rarely have to include this file directly).</p>
				349
				350	<dl>
				351	<dt><tt>isa<></tt>: </dt>
				352
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	353	<dd><p>The <tt>isa<></tt> operator works exactly like the Java
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	354	"<tt>instanceof</tt>" operator. It returns true or false depending on whether
				355	a reference or pointer points to an instance of the specified class. This can
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	356	be very useful for constraint checking of various sorts (example below).</p>
				357	</dd>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	358
				359	<dt><tt>cast<></tt>: </dt>
				360
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	361	<dd><p>The <tt>cast<></tt> operator is a "checked cast" operation. It
Chris Lattner	28e6ff5	2008-06-20 05:03:17 +0000	[diff] [blame]	362	converts a pointer or reference from a base class to a derived class, causing
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	363	an assertion failure if it is not really an instance of the right type. This
				364	should be used in cases where you have some information that makes you believe
				365	that something is of the right type. An example of the <tt>isa<></tt>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	366	and <tt>cast<></tt> template is:</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	367
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	368	<div class="doc_code">
				369	<pre>
				370	static bool isLoopInvariant(const <a href="#Value">Value</a> V, const Loop L) {
				371	if (isa<<a href="#Constant">Constant</a>>(V) \|\| isa<<a href="#Argument">Argument</a>>(V) \|\| isa<<a href="#GlobalValue">GlobalValue</a>>(V))
				372	return true;
Chris Lattner	69bf8a9	2004-05-23 21:06:58 +0000	[diff] [blame]	373
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	374	// <i>Otherwise, it must be an instruction...</i>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	375	return !L->contains(cast<<a href="#Instruction">Instruction</a>>(V)->getParent());
				376	}
				377	</pre>
				378	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	379
				380	<p>Note that you should <b>not</b> use an <tt>isa<></tt> test followed
				381	by a <tt>cast<></tt>, for that use the <tt>dyn_cast<></tt>
				382	operator.</p>
				383
				384	</dd>
				385
				386	<dt><tt>dyn_cast<></tt>:</dt>
				387
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	388	<dd><p>The <tt>dyn_cast<></tt> operator is a "checking cast" operation.
				389	It checks to see if the operand is of the specified type, and if so, returns a
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	390	pointer to it (this operator does not work with references). If the operand is
				391	not of the correct type, a null pointer is returned. Thus, this works very
Misha Brukman	2c122ce	2005-11-01 21:12:49 +0000	[diff] [blame]	392	much like the <tt>dynamic_cast<></tt> operator in C++, and should be
				393	used in the same circumstances. Typically, the <tt>dyn_cast<></tt>
				394	operator is used in an <tt>if</tt> statement or some other flow control
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	395	statement like this:</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	396
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	397	<div class="doc_code">
				398	<pre>
				399	if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast<<a href="#AllocationInst">AllocationInst</a>>(Val)) {
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	400	// <i>...</i>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	401	}
				402	</pre>
				403	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	404
Misha Brukman	2c122ce	2005-11-01 21:12:49 +0000	[diff] [blame]	405	<p>This form of the <tt>if</tt> statement effectively combines together a call
				406	to <tt>isa<></tt> and a call to <tt>cast<></tt> into one
				407	statement, which is very convenient.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	408
Misha Brukman	2c122ce	2005-11-01 21:12:49 +0000	[diff] [blame]	409	<p>Note that the <tt>dyn_cast<></tt> operator, like C++'s
				410	<tt>dynamic_cast<></tt> or Java's <tt>instanceof</tt> operator, can be
				411	abused. In particular, you should not use big chained <tt>if/then/else</tt>
				412	blocks to check for lots of different variants of classes. If you find
				413	yourself wanting to do this, it is much cleaner and more efficient to use the
				414	<tt>InstVisitor</tt> class to dispatch over the instruction type directly.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	415
Misha Brukman	2c122ce	2005-11-01 21:12:49 +0000	[diff] [blame]	416	</dd>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	417
Misha Brukman	2c122ce	2005-11-01 21:12:49 +0000	[diff] [blame]	418	<dt><tt>cast_or_null<></tt>: </dt>
				419
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	420	<dd><p>The <tt>cast_or_null<></tt> operator works just like the
Misha Brukman	2c122ce	2005-11-01 21:12:49 +0000	[diff] [blame]	421	<tt>cast<></tt> operator, except that it allows for a null pointer as an
				422	argument (which it then propagates). This can sometimes be useful, allowing
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	423	you to combine several null checks into one.</p></dd>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	424
Misha Brukman	2c122ce	2005-11-01 21:12:49 +0000	[diff] [blame]	425	<dt><tt>dyn_cast_or_null<></tt>: </dt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	426
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	427	<dd><p>The <tt>dyn_cast_or_null<></tt> operator works just like the
Misha Brukman	2c122ce	2005-11-01 21:12:49 +0000	[diff] [blame]	428	<tt>dyn_cast<></tt> operator, except that it allows for a null pointer
				429	as an argument (which it then propagates). This can sometimes be useful,
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	430	allowing you to combine several null checks into one.</p></dd>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	431
Misha Brukman	2c122ce	2005-11-01 21:12:49 +0000	[diff] [blame]	432	</dl>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	433
				434	<p>These five templates can be used with any classes, whether they have a
				435	v-table or not. To add support for these templates, you simply need to add
				436	<tt>classof</tt> static methods to the class you are interested casting
				437	to. Describing this is currently outside the scope of this document, but there
				438	are lots of examples in the LLVM source base.</p>
				439
				440	</div>
				441
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	442
				443	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	444	<h3>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	445	<a name="string_apis">Passing strings (the <tt>StringRef</tt>
				446	and <tt>Twine</tt> classes)</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	447	</h3>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	448
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	449	<div>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	450
				451	<p>Although LLVM generally does not do much string manipulation, we do have
Chris Lattner	81187ae	2009-07-25 07:16:59 +0000	[diff] [blame]	452	several important APIs which take strings. Two important examples are the
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	453	Value class -- which has names for instructions, functions, etc. -- and the
				454	StringMap class which is used extensively in LLVM and Clang.</p>
				455
				456	<p>These are generic classes, and they need to be able to accept strings which
				457	may have embedded null characters. Therefore, they cannot simply take
Chris Lattner	81187ae	2009-07-25 07:16:59 +0000	[diff] [blame]	458	a <tt>const char *</tt>, and taking a <tt>const std::string&</tt> requires
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	459	clients to perform a heap allocation which is usually unnecessary. Instead,
Benjamin Kramer	38e5989	2010-07-14 22:38:02 +0000	[diff] [blame]	460	many LLVM APIs use a <tt>StringRef</tt> or a <tt>const Twine&</tt> for
				461	passing strings efficiently.</p>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	462
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	463	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	464	<h4>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	465	<a name="StringRef">The <tt>StringRef</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	466	</h4>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	467
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	468	<div>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	469
				470	<p>The <tt>StringRef</tt> data type represents a reference to a constant string
				471	(a character array and a length) and supports the common operations available
				472	on <tt>std:string</tt>, but does not require heap allocation.</p>
				473
Chris Lattner	81187ae	2009-07-25 07:16:59 +0000	[diff] [blame]	474	<p>It can be implicitly constructed using a C style null-terminated string,
				475	an <tt>std::string</tt>, or explicitly with a character pointer and length.
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	476	For example, the <tt>StringRef</tt> find function is declared as:</p>
Chris Lattner	81187ae	2009-07-25 07:16:59 +0000	[diff] [blame]	477
Benjamin Kramer	38e5989	2010-07-14 22:38:02 +0000	[diff] [blame]	478	<pre class="doc_code">
				479	iterator find(StringRef Key);
				480	</pre>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	481
				482	<p>and clients can call it using any one of:</p>
				483
Benjamin Kramer	38e5989	2010-07-14 22:38:02 +0000	[diff] [blame]	484	<pre class="doc_code">
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	485	Map.find("foo"); <i>// Lookup "foo"</i>
				486	Map.find(std::string("bar")); <i>// Lookup "bar"</i>
				487	Map.find(StringRef("\0baz", 4)); <i>// Lookup "\0baz"</i>
				488	</pre>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	489
				490	<p>Similarly, APIs which need to return a string may return a <tt>StringRef</tt>
				491	instance, which can be used directly or converted to an <tt>std::string</tt>
				492	using the <tt>str</tt> member function. See
				493	"<tt><a href="/doxygen/classllvm_1_1StringRef_8h-source.html">llvm/ADT/StringRef.h</a></tt>"
				494	for more information.</p>
				495
				496	<p>You should rarely use the <tt>StringRef</tt> class directly, because it contains
				497	pointers to external memory it is not generally safe to store an instance of the
Benjamin Kramer	38e5989	2010-07-14 22:38:02 +0000	[diff] [blame]	498	class (unless you know that the external storage will not be freed). StringRef is
				499	small and pervasive enough in LLVM that it should always be passed by value.</p>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	500
				501	</div>
				502
				503	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	504	<h4>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	505	<a name="Twine">The <tt>Twine</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	506	</h4>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	507
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	508	<div>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	509
Axel Naumann	fc97547	2012-07-25 13:46:11 +0000	[diff] [blame^]	510	<p>The <tt><a href="/doxygen/classllvm_1_1Twine.html">Twine</a></tt> class is an
				511	efficient way for APIs to accept concatenated strings. For example, a common
				512	LLVM paradigm is to name one instruction based on
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	513	the name of another instruction with a suffix, for example:</p>
				514
				515	<div class="doc_code">
				516	<pre>
				517	New = CmpInst::Create(<i>...</i>, SO->getName() + ".cmp");
				518	</pre>
				519	</div>
				520
Axel Naumann	fc97547	2012-07-25 13:46:11 +0000	[diff] [blame^]	521	<p>The <tt>Twine</tt> class is effectively a lightweight
				522	<a href="http://en.wikipedia.org/wiki/Rope_(computer_science)">rope</a>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	523	which points to temporary (stack allocated) objects. Twines can be implicitly
				524	constructed as the result of the plus operator applied to strings (i.e., a C
Axel Naumann	fc97547	2012-07-25 13:46:11 +0000	[diff] [blame^]	525	strings, an <tt>std::string</tt>, or a <tt>StringRef</tt>). The twine delays
				526	the actual concatenation of strings until it is actually required, at which
				527	point it can be efficiently rendered directly into a character array. This
				528	avoids unnecessary heap allocation involved in constructing the temporary
				529	results of string concatenation. See
				530	"<tt><a href="/doxygen/Twine_8h_source.html">llvm/ADT/Twine.h</a></tt>"
				531	and <a href="#dss_twine">here</a> for more information.</p>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	532
				533	<p>As with a <tt>StringRef</tt>, <tt>Twine</tt> objects point to external memory
				534	and should almost never be stored or mentioned directly. They are intended
				535	solely for use when defining a function which should be able to efficiently
				536	accept concatenated strings.</p>
				537
				538	</div>
				539
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	540	</div>
Daniel Dunbar	6e0d1cb	2009-07-25 04:41:11 +0000	[diff] [blame]	541
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	542	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	543	<h3>
Misha Brukman	2c122ce	2005-11-01 21:12:49 +0000	[diff] [blame]	544	<a name="DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt> option</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	545	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	546
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	547	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	548
				549	<p>Often when working on your pass you will put a bunch of debugging printouts
				550	and other code into your pass. After you get it working, you want to remove
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	551	it, but you may need it again in the future (to work out new bugs that you run
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	552	across).</p>
				553
				554	<p> Naturally, because of this, you don't want to delete the debug printouts,
				555	but you don't want them to always be noisy. A standard compromise is to comment
				556	them out, allowing you to enable them if you need them in the future.</p>
				557
Chris Lattner	695b78b	2005-04-26 22:56:16 +0000	[diff] [blame]	558	<p>The "<tt><a href="/doxygen/Debug_8h-source.html">llvm/Support/Debug.h</a></tt>"
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	559	file provides a macro named <tt>DEBUG()</tt> that is a much nicer solution to
				560	this problem. Basically, you can put arbitrary code into the argument of the
				561	<tt>DEBUG</tt> macro, and it is only executed if '<tt>opt</tt>' (or any other
				562	tool) is run with the '<tt>-debug</tt>' command line argument:</p>
				563
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	564	<div class="doc_code">
				565	<pre>
Daniel Dunbar	06388ae	2009-07-25 01:55:32 +0000	[diff] [blame]	566	DEBUG(errs() << "I am here!\n");
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	567	</pre>
				568	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	569
				570	<p>Then you can run your pass like this:</p>
				571
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	572	<div class="doc_code">
				573	<pre>
				574	$ opt < a.bc > /dev/null -mypass
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	575	<i><no output></i>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	576	$ opt < a.bc > /dev/null -mypass -debug
				577	I am here!
				578	</pre>
				579	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	580
				581	<p>Using the <tt>DEBUG()</tt> macro instead of a home-brewed solution allows you
				582	to not have to create "yet another" command line option for the debug output for
				583	your pass. Note that <tt>DEBUG()</tt> macros are disabled for optimized builds,
				584	so they do not cause a performance impact at all (for the same reason, they
				585	should also not contain side-effects!).</p>
				586
				587	<p>One additional nice thing about the <tt>DEBUG()</tt> macro is that you can
				588	enable or disable it directly in gdb. Just use "<tt>set DebugFlag=0</tt>" or
				589	"<tt>set DebugFlag=1</tt>" from the gdb if the program is running. If the
				590	program hasn't been started yet, you can always just run it with
				591	<tt>-debug</tt>.</p>
				592
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	593	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	594	<h4>
Chris Lattner	c915108	2005-04-26 22:57:07 +0000	[diff] [blame]	595	<a name="DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt> and
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	596	the <tt>-debug-only</tt> option</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	597	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	598
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	599	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	600
				601	<p>Sometimes you may find yourself in a situation where enabling <tt>-debug</tt>
				602	just turns on <b>too much</b> information (such as when working on the code
				603	generator). If you want to enable debug information with more fine-grained
				604	control, you define the <tt>DEBUG_TYPE</tt> macro and the <tt>-debug</tt> only
				605	option as follows:</p>
				606
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	607	<div class="doc_code">
				608	<pre>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	609	#undef DEBUG_TYPE
Daniel Dunbar	06388ae	2009-07-25 01:55:32 +0000	[diff] [blame]	610	DEBUG(errs() << "No debug type\n");
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	611	#define DEBUG_TYPE "foo"
Daniel Dunbar	06388ae	2009-07-25 01:55:32 +0000	[diff] [blame]	612	DEBUG(errs() << "'foo' debug type\n");
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	613	#undef DEBUG_TYPE
				614	#define DEBUG_TYPE "bar"
Daniel Dunbar	06388ae	2009-07-25 01:55:32 +0000	[diff] [blame]	615	DEBUG(errs() << "'bar' debug type\n"));
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	616	#undef DEBUG_TYPE
				617	#define DEBUG_TYPE ""
Daniel Dunbar	06388ae	2009-07-25 01:55:32 +0000	[diff] [blame]	618	DEBUG(errs() << "No debug type (2)\n");
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	619	</pre>
				620	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	621
				622	<p>Then you can run your pass like this:</p>
				623
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	624	<div class="doc_code">
				625	<pre>
				626	$ opt < a.bc > /dev/null -mypass
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	627	<i><no output></i>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	628	$ opt < a.bc > /dev/null -mypass -debug
				629	No debug type
				630	'foo' debug type
				631	'bar' debug type
				632	No debug type (2)
				633	$ opt < a.bc > /dev/null -mypass -debug-only=foo
				634	'foo' debug type
				635	$ opt < a.bc > /dev/null -mypass -debug-only=bar
				636	'bar' debug type
				637	</pre>
				638	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	639
				640	<p>Of course, in practice, you should only set <tt>DEBUG_TYPE</tt> at the top of
				641	a file, to specify the debug type for the entire module (if you do this before
Chris Lattner	695b78b	2005-04-26 22:56:16 +0000	[diff] [blame]	642	you <tt>#include "llvm/Support/Debug.h"</tt>, you don't have to insert the ugly
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	643	<tt>#undef</tt>'s). Also, you should use names more meaningful than "foo" and
				644	"bar", because there is no system in place to ensure that names do not
				645	conflict. If two different modules use the same string, they will all be turned
				646	on when the name is specified. This allows, for example, all debug information
				647	for instruction scheduling to be enabled with <tt>-debug-type=InstrSched</tt>,
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	648	even if the source lives in multiple files.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	649
Daniel Dunbar	c3c9239	2009-08-07 23:48:59 +0000	[diff] [blame]	650	<p>The <tt>DEBUG_WITH_TYPE</tt> macro is also available for situations where you
				651	would like to set <tt>DEBUG_TYPE</tt>, but only for one specific <tt>DEBUG</tt>
				652	statement. It takes an additional first parameter, which is the type to use. For
Benjamin Kramer	8040cd3	2009-10-12 14:46:08 +0000	[diff] [blame]	653	example, the preceding example could be written as:</p>
Daniel Dunbar	c3c9239	2009-08-07 23:48:59 +0000	[diff] [blame]	654
				655
				656	<div class="doc_code">
				657	<pre>
				658	DEBUG_WITH_TYPE("", errs() << "No debug type\n");
				659	DEBUG_WITH_TYPE("foo", errs() << "'foo' debug type\n");
				660	DEBUG_WITH_TYPE("bar", errs() << "'bar' debug type\n"));
				661	DEBUG_WITH_TYPE("", errs() << "No debug type (2)\n");
				662	</pre>
				663	</div>
				664
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	665	</div>
				666
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	667	</div>
				668
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	669	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	670	<h3>
Chris Lattner	0be6fdf	2006-12-19 21:46:21 +0000	[diff] [blame]	671	<a name="Statistic">The <tt>Statistic</tt> class & <tt>-stats</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	672	option</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	673	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	674
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	675	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	676
				677	<p>The "<tt><a
Chris Lattner	695b78b	2005-04-26 22:56:16 +0000	[diff] [blame]	678	href="/doxygen/Statistic_8h-source.html">llvm/ADT/Statistic.h</a></tt>" file
Chris Lattner	0be6fdf	2006-12-19 21:46:21 +0000	[diff] [blame]	679	provides a class named <tt>Statistic</tt> that is used as a unified way to
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	680	keep track of what the LLVM compiler is doing and how effective various
				681	optimizations are. It is useful to see what optimizations are contributing to
				682	making a particular program run faster.</p>
				683
				684	<p>Often you may run your pass on some big program, and you're interested to see
				685	how many times it makes a certain transformation. Although you can do this with
				686	hand inspection, or some ad-hoc method, this is a real pain and not very useful
Chris Lattner	0be6fdf	2006-12-19 21:46:21 +0000	[diff] [blame]	687	for big programs. Using the <tt>Statistic</tt> class makes it very easy to
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	688	keep track of this information, and the calculated information is presented in a
				689	uniform manner with the rest of the passes being executed.</p>
				690
				691	<p>There are many examples of <tt>Statistic</tt> uses, but the basics of using
				692	it are as follows:</p>
				693
				694	<ol>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	695	<li><p>Define your statistic like this:</p>
				696
				697	<div class="doc_code">
				698	<pre>
Chris Lattner	0be6fdf	2006-12-19 21:46:21 +0000	[diff] [blame]	699	#define <a href="#DEBUG_TYPE">DEBUG_TYPE</a> "mypassname" <i>// This goes before any #includes.</i>
				700	STATISTIC(NumXForms, "The # of times I did stuff");
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	701	</pre>
				702	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	703
Chris Lattner	0be6fdf	2006-12-19 21:46:21 +0000	[diff] [blame]	704	<p>The <tt>STATISTIC</tt> macro defines a static variable, whose name is
				705	specified by the first argument. The pass name is taken from the DEBUG_TYPE
				706	macro, and the description is taken from the second argument. The variable
Reid Spencer	06565dc	2007-01-12 17:11:23 +0000	[diff] [blame]	707	defined ("NumXForms" in this case) acts like an unsigned integer.</p></li>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	708
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	709	<li><p>Whenever you make a transformation, bump the counter:</p>
				710
				711	<div class="doc_code">
				712	<pre>
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	713	++NumXForms; // <i>I did stuff!</i>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	714	</pre>
				715	</div>
				716
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	717	</li>
				718	</ol>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	719
				720	<p>That's all you have to do. To get '<tt>opt</tt>' to print out the
				721	statistics gathered, use the '<tt>-stats</tt>' option:</p>
				722
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	723	<div class="doc_code">
				724	<pre>
				725	$ opt -stats -mypassname < program.bc > /dev/null
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	726	<i>... statistics output ...</i>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	727	</pre>
				728	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	729
Reid Spencer	6b6c73e	2007-02-09 16:00:28 +0000	[diff] [blame]	730	<p> When running <tt>opt</tt> on a C file from the SPEC benchmark
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	731	suite, it gives a report that looks like this:</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	732
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	733	<div class="doc_code">
				734	<pre>
Gabor Greif	04367bf	2007-07-06 22:07:22 +0000	[diff] [blame]	735	7646 bitcodewriter - Number of normal instructions
				736	725 bitcodewriter - Number of oversized instructions
				737	129996 bitcodewriter - Number of bitcode bytes written
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	738	2817 raise - Number of insts DCEd or constprop'd
				739	3213 raise - Number of cast-of-self removed
				740	5046 raise - Number of expression trees converted
				741	75 raise - Number of other getelementptr's formed
				742	138 raise - Number of load/store peepholes
				743	42 deadtypeelim - Number of unused typenames removed from symtab
				744	392 funcresolve - Number of varargs functions resolved
				745	27 globaldce - Number of global variables removed
				746	2 adce - Number of basic blocks removed
				747	134 cee - Number of branches revectored
				748	49 cee - Number of setcc instruction eliminated
				749	532 gcse - Number of loads removed
				750	2919 gcse - Number of instructions removed
				751	86 indvars - Number of canonical indvars added
				752	87 indvars - Number of aux indvars removed
				753	25 instcombine - Number of dead inst eliminate
				754	434 instcombine - Number of insts combined
				755	248 licm - Number of load insts hoisted
				756	1298 licm - Number of insts hoisted to a loop pre-header
				757	3 licm - Number of insts hoisted to multiple loop preds (bad, no loop pre-header)
				758	75 mem2reg - Number of alloca's promoted
				759	1444 cfgsimplify - Number of blocks simplified
				760	</pre>
				761	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	762
				763	<p>Obviously, with so many optimizations, having a unified framework for this
				764	stuff is very nice. Making your pass fit well into the framework makes it more
				765	maintainable and useful.</p>
				766
				767	</div>
				768
Chris Lattner	f623a08	2005-10-17 01:36:23 +0000	[diff] [blame]	769	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	770	<h3>
Chris Lattner	f623a08	2005-10-17 01:36:23 +0000	[diff] [blame]	771	<a name="ViewGraph">Viewing graphs while debugging code</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	772	</h3>
Chris Lattner	f623a08	2005-10-17 01:36:23 +0000	[diff] [blame]	773
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	774	<div>
Chris Lattner	f623a08	2005-10-17 01:36:23 +0000	[diff] [blame]	775
				776	<p>Several of the important data structures in LLVM are graphs: for example
				777	CFGs made out of LLVM <a href="#BasicBlock">BasicBlock</a>s, CFGs made out of
				778	LLVM <a href="CodeGenerator.html#machinebasicblock">MachineBasicBlock</a>s, and
				779	<a href="CodeGenerator.html#selectiondag_intro">Instruction Selection
				780	DAGs</a>. In many cases, while debugging various parts of the compiler, it is
				781	nice to instantly visualize these graphs.</p>
				782
				783	<p>LLVM provides several callbacks that are available in a debug build to do
				784	exactly that. If you call the <tt>Function::viewCFG()</tt> method, for example,
				785	the current LLVM tool will pop up a window containing the CFG for the function
				786	where each basic block is a node in the graph, and each node contains the
				787	instructions in the block. Similarly, there also exists
				788	<tt>Function::viewCFGOnly()</tt> (does not include the instructions), the
				789	<tt>MachineFunction::viewCFG()</tt> and <tt>MachineFunction::viewCFGOnly()</tt>,
				790	and the <tt>SelectionDAG::viewGraph()</tt> methods. Within GDB, for example,
Jim Laskey	543a0ee	2006-10-02 12:28:07 +0000	[diff] [blame]	791	you can usually use something like <tt>call DAG.viewGraph()</tt> to pop
Chris Lattner	f623a08	2005-10-17 01:36:23 +0000	[diff] [blame]	792	up a window. Alternatively, you can sprinkle calls to these functions in your
				793	code in places you want to debug.</p>
				794
				795	<p>Getting this to work requires a small amount of configuration. On Unix
				796	systems with X11, install the <a href="http://www.graphviz.org">graphviz</a>
				797	toolkit, and make sure 'dot' and 'gv' are in your path. If you are running on
				798	Mac OS/X, download and install the Mac OS/X <a
				799	href="http://www.pixelglow.com/graphviz/">Graphviz program</a>, and add
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	800	<tt>/Applications/Graphviz.app/Contents/MacOS/</tt> (or wherever you install
Chris Lattner	f623a08	2005-10-17 01:36:23 +0000	[diff] [blame]	801	it) to your path. Once in your system and path are set up, rerun the LLVM
				802	configure script and rebuild LLVM to enable this functionality.</p>
				803
Jim Laskey	543a0ee	2006-10-02 12:28:07 +0000	[diff] [blame]	804	<p><tt>SelectionDAG</tt> has been extended to make it easier to locate
				805	<i>interesting</i> nodes in large complex graphs. From gdb, if you
				806	<tt>call DAG.setGraphColor(<i>node</i>, "<i>color</i>")</tt>, then the
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	807	next <tt>call DAG.viewGraph()</tt> would highlight the node in the
Jim Laskey	543a0ee	2006-10-02 12:28:07 +0000	[diff] [blame]	808	specified color (choices of colors can be found at <a
Chris Lattner	302da1e	2007-02-03 03:05:57 +0000	[diff] [blame]	809	href="http://www.graphviz.org/doc/info/colors.html">colors</a>.) More
Jim Laskey	543a0ee	2006-10-02 12:28:07 +0000	[diff] [blame]	810	complex node attributes can be provided with <tt>call
				811	DAG.setGraphAttrs(<i>node</i>, "<i>attributes</i>")</tt> (choices can be
				812	found at <a href="http://www.graphviz.org/doc/info/attrs.html">Graph
				813	Attributes</a>.) If you want to restart and clear all the current graph
				814	attributes, then you can <tt>call DAG.clearGraphAttrs()</tt>. </p>
				815
Chris Lattner	83f9467	2011-06-13 15:59:35 +0000	[diff] [blame]	816	<p>Note that graph visualization features are compiled out of Release builds
				817	to reduce file size. This means that you need a Debug+Asserts or
				818	Release+Asserts build to use these features.</p>
				819
Chris Lattner	f623a08	2005-10-17 01:36:23 +0000	[diff] [blame]	820	</div>
				821
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	822	</div>
				823
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	824	<!-- *********************************************************************** -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	825	<h2>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	826	<a name="datastructure">Picking the Right Data Structure for a Task</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	827	</h2>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	828	<!-- *********************************************************************** -->
				829
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	830	<div>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	831
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	832	<p>LLVM has a plethora of data structures in the <tt>llvm/ADT/</tt> directory,
				833	and we commonly use STL data structures. This section describes the trade-offs
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	834	you should consider when you pick one.</p>
				835
				836	<p>
				837	The first step is a choose your own adventure: do you want a sequential
				838	container, a set-like container, or a map-like container? The most important
				839	thing when choosing a container is the algorithmic properties of how you plan to
				840	access the container. Based on that, you should use:</p>
				841
				842	<ul>
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	843	<li>a <a href="#ds_map">map-like</a> container if you need efficient look-up
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	844	of an value based on another value. Map-like containers also support
				845	efficient queries for containment (whether a key is in the map). Map-like
				846	containers generally do not support efficient reverse mapping (values to
				847	keys). If you need that, use two maps. Some map-like containers also
				848	support efficient iteration through the keys in sorted order. Map-like
				849	containers are the most expensive sort, only use them if you need one of
				850	these capabilities.</li>
				851
				852	<li>a <a href="#ds_set">set-like</a> container if you need to put a bunch of
				853	stuff into a container that automatically eliminates duplicates. Some
				854	set-like containers support efficient iteration through the elements in
				855	sorted order. Set-like containers are more expensive than sequential
				856	containers.
				857	</li>
				858
				859	<li>a <a href="#ds_sequential">sequential</a> container provides
				860	the most efficient way to add elements and keeps track of the order they are
				861	added to the collection. They permit duplicates and support efficient
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	862	iteration, but do not support efficient look-up based on a key.
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	863	</li>
				864
Chris Lattner	dced9fb	2009-07-25 07:22:20 +0000	[diff] [blame]	865	<li>a <a href="#ds_string">string</a> container is a specialized sequential
				866	container or reference structure that is used for character or byte
				867	arrays.</li>
				868
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	869	<li>a <a href="#ds_bit">bit</a> container provides an efficient way to store and
				870	perform set operations on sets of numeric id's, while automatically
				871	eliminating duplicates. Bit containers require a maximum of 1 bit for each
				872	identifier you want to store.
				873	</li>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	874	</ul>
				875
				876	<p>
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	877	Once the proper category of container is determined, you can fine tune the
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	878	memory use, constant factors, and cache behaviors of access by intelligently
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	879	picking a member of the category. Note that constant factors and cache behavior
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	880	can be a big deal. If you have a vector that usually only contains a few
				881	elements (but could contain many), for example, it's much better to use
				882	<a href="#dss_smallvector">SmallVector</a> than <a href="#dss_vector">vector</a>
				883	. Doing so avoids (relatively) expensive malloc/free calls, which dwarf the
				884	cost of adding the elements to the container. </p>
				885
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	886	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	887	<h3>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	888	<a name="ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	889	</h3>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	890
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	891	<div>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	892	There are a variety of sequential containers available for you, based on your
				893	needs. Pick the first in this section that will do what you want.
Chris Lattner	3b4f417	2011-07-22 21:34:12 +0000	[diff] [blame]	894
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	895	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	896	<h4>
Chris Lattner	8ae4261	2011-04-05 23:18:20 +0000	[diff] [blame]	897	<a name="dss_arrayref">llvm/ADT/ArrayRef.h</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	898	</h4>
Chris Lattner	8ae4261	2011-04-05 23:18:20 +0000	[diff] [blame]	899
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	900	<div>
Chris Lattner	8ae4261	2011-04-05 23:18:20 +0000	[diff] [blame]	901	<p>The llvm::ArrayRef class is the preferred class to use in an interface that
				902	accepts a sequential list of elements in memory and just reads from them. By
				903	taking an ArrayRef, the API can be passed a fixed size array, an std::vector,
				904	an llvm::SmallVector and anything else that is contiguous in memory.
				905	</p>
				906	</div>
				907
				908
				909
				910	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	911	<h4>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	912	<a name="dss_fixedarrays">Fixed Size Arrays</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	913	</h4>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	914
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	915	<div>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	916	<p>Fixed size arrays are very simple and very fast. They are good if you know
				917	exactly how many elements you have, or you have a (low) upper bound on how many
				918	you have.</p>
				919	</div>
				920
				921	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	922	<h4>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	923	<a name="dss_heaparrays">Heap Allocated Arrays</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	924	</h4>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	925
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	926	<div>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	927	<p>Heap allocated arrays (new[] + delete[]) are also simple. They are good if
				928	the number of elements is variable, if you know how many elements you will need
				929	before the array is allocated, and if the array is usually large (if not,
				930	consider a <a href="#dss_smallvector">SmallVector</a>). The cost of a heap
				931	allocated array is the cost of the new/delete (aka malloc/free). Also note that
				932	if you are allocating an array of a type with a constructor, the constructor and
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	933	destructors will be run for every element in the array (re-sizable vectors only
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	934	construct those elements actually used).</p>
				935	</div>
				936
				937	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	938	<h4>
Chris Lattner	9d69d4a	2011-07-18 01:40:02 +0000	[diff] [blame]	939	<a name="dss_tinyptrvector">"llvm/ADT/TinyPtrVector.h"</a>
				940	</h4>
				941
				942
				943	<div>
				944	<p><tt>TinyPtrVector<Type></tt> is a highly specialized collection class
				945	that is optimized to avoid allocation in the case when a vector has zero or one
				946	elements. It has two major restrictions: 1) it can only hold values of pointer
				947	type, and 2) it cannot hold a null pointer.</p>
				948
				949	<p>Since this container is highly specialized, it is rarely used.</p>
				950
				951	</div>
				952
Chris Lattner	9d69d4a	2011-07-18 01:40:02 +0000	[diff] [blame]	953	<!-- _______________________________________________________________________ -->
				954	<h4>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	955	<a name="dss_smallvector">"llvm/ADT/SmallVector.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	956	</h4>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	957
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	958	<div>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	959	<p><tt>SmallVector<Type, N></tt> is a simple class that looks and smells
				960	just like <tt>vector<Type></tt>:
				961	it supports efficient iteration, lays out elements in memory order (so you can
				962	do pointer arithmetic between elements), supports efficient push_back/pop_back
				963	operations, supports efficient random access to its elements, etc.</p>
				964
				965	<p>The advantage of SmallVector is that it allocates space for
				966	some number of elements (N) <b>in the object itself</b>. Because of this, if
				967	the SmallVector is dynamically smaller than N, no malloc is performed. This can
				968	be a big win in cases where the malloc/free call is far more expensive than the
				969	code that fiddles around with the elements.</p>
				970
				971	<p>This is good for vectors that are "usually small" (e.g. the number of
				972	predecessors/successors of a block is usually less than 8). On the other hand,
				973	this makes the size of the SmallVector itself large, so you don't want to
				974	allocate lots of them (doing so will waste a lot of space). As such,
				975	SmallVectors are most useful when on the stack.</p>
				976
				977	<p>SmallVector also provides a nice portable and efficient replacement for
				978	<tt>alloca</tt>.</p>
				979
				980	</div>
				981
				982	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	983	<h4>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	984	<a name="dss_vector"><vector></a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	985	</h4>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	986
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	987	<div>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	988	<p>
				989	std::vector is well loved and respected. It is useful when SmallVector isn't:
				990	when the size of the vector is often large (thus the small optimization will
				991	rarely be a benefit) or if you will be allocating many instances of the vector
				992	itself (which would waste space for elements that aren't in the container).
				993	vector is also useful when interfacing with code that expects vectors :).
				994	</p>
Chris Lattner	32d8476	2007-02-05 06:30:51 +0000	[diff] [blame]	995
				996	<p>One worthwhile note about std::vector: avoid code like this:</p>
				997
				998	<div class="doc_code">
				999	<pre>
				1000	for ( ... ) {
Chris Lattner	9bb3dbb	2007-03-28 18:27:57 +0000	[diff] [blame]	1001	std::vector<foo> V;
Jim Grosbach	087f050	2011-10-28 20:52:20 +0000	[diff] [blame]	1002	// make use of V.
Chris Lattner	32d8476	2007-02-05 06:30:51 +0000	[diff] [blame]	1003	}
				1004	</pre>
				1005	</div>
				1006
				1007	<p>Instead, write this as:</p>
				1008
				1009	<div class="doc_code">
				1010	<pre>
Chris Lattner	9bb3dbb	2007-03-28 18:27:57 +0000	[diff] [blame]	1011	std::vector<foo> V;
Chris Lattner	32d8476	2007-02-05 06:30:51 +0000	[diff] [blame]	1012	for ( ... ) {
Jim Grosbach	087f050	2011-10-28 20:52:20 +0000	[diff] [blame]	1013	// make use of V.
Chris Lattner	32d8476	2007-02-05 06:30:51 +0000	[diff] [blame]	1014	V.clear();
				1015	}
				1016	</pre>
				1017	</div>
				1018
				1019	<p>Doing so will save (at least) one heap allocation and free per iteration of
				1020	the loop.</p>
				1021
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1022	</div>
				1023
				1024	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1025	<h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1026	<a name="dss_deque"><deque></a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1027	</h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1028
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1029	<div>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1030	<p>std::deque is, in some senses, a generalized version of std::vector. Like
				1031	std::vector, it provides constant time random access and other similar
				1032	properties, but it also provides efficient access to the front of the list. It
				1033	does not guarantee continuity of elements within memory.</p>
				1034
				1035	<p>In exchange for this extra flexibility, std::deque has significantly higher
				1036	constant factor costs than std::vector. If possible, use std::vector or
				1037	something cheaper.</p>
				1038	</div>
				1039
				1040	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1041	<h4>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1042	<a name="dss_list"><list></a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1043	</h4>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1044
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1045	<div>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1046	<p>std::list is an extremely inefficient class that is rarely useful.
				1047	It performs a heap allocation for every element inserted into it, thus having an
				1048	extremely high constant factor, particularly for small data types. std::list
				1049	also only supports bidirectional iteration, not random access iteration.</p>
				1050
				1051	<p>In exchange for this high cost, std::list supports efficient access to both
				1052	ends of the list (like std::deque, but unlike std::vector or SmallVector). In
				1053	addition, the iterator invalidation characteristics of std::list are stronger
				1054	than that of a vector class: inserting or removing an element into the list does
				1055	not invalidate iterator or pointers to other elements in the list.</p>
				1056	</div>
				1057
				1058	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1059	<h4>
Gabor Greif	3899e49	2009-02-27 11:37:41 +0000	[diff] [blame]	1060	<a name="dss_ilist">llvm/ADT/ilist.h</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1061	</h4>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1062
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1063	<div>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1064	<p><tt>ilist<T></tt> implements an 'intrusive' doubly-linked list. It is
				1065	intrusive, because it requires the element to store and provide access to the
				1066	prev/next pointers for the list.</p>
				1067
Gabor Greif	2946d1c	2009-02-27 12:02:19 +0000	[diff] [blame]	1068	<p><tt>ilist</tt> has the same drawbacks as <tt>std::list</tt>, and additionally
				1069	requires an <tt>ilist_traits</tt> implementation for the element type, but it
				1070	provides some novel characteristics. In particular, it can efficiently store
				1071	polymorphic objects, the traits class is informed when an element is inserted or
Gabor Greif	0cbcabe	2009-03-12 09:47:03 +0000	[diff] [blame]	1072	removed from the list, and <tt>ilist</tt>s are guaranteed to support a
				1073	constant-time splice operation.</p>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1074
Gabor Greif	0cbcabe	2009-03-12 09:47:03 +0000	[diff] [blame]	1075	<p>These properties are exactly what we want for things like
				1076	<tt>Instruction</tt>s and basic blocks, which is why these are implemented with
				1077	<tt>ilist</tt>s.</p>
Gabor Greif	3899e49	2009-02-27 11:37:41 +0000	[diff] [blame]	1078
				1079	Related classes of interest are explained in the following subsections:
				1080	<ul>
Gabor Greif	0186250	2009-02-27 13:28:07 +0000	[diff] [blame]	1081	<li><a href="#dss_ilist_traits">ilist_traits</a></li>
Gabor Greif	2946d1c	2009-02-27 12:02:19 +0000	[diff] [blame]	1082	<li><a href="#dss_iplist">iplist</a></li>
Gabor Greif	3899e49	2009-02-27 11:37:41 +0000	[diff] [blame]	1083	<li><a href="#dss_ilist_node">llvm/ADT/ilist_node.h</a></li>
Gabor Greif	6a65f42	2009-03-12 10:30:31 +0000	[diff] [blame]	1084	<li><a href="#dss_ilist_sentinel">Sentinels</a></li>
Gabor Greif	3899e49	2009-02-27 11:37:41 +0000	[diff] [blame]	1085	</ul>
				1086	</div>
				1087
				1088	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1089	<h4>
Argyrios Kyrtzidis	81df089	2011-06-15 19:56:01 +0000	[diff] [blame]	1090	<a name="dss_packedvector">llvm/ADT/PackedVector.h</a>
				1091	</h4>
				1092
				1093	<div>
				1094	<p>
				1095	Useful for storing a vector of values using only a few number of bits for each
				1096	value. Apart from the standard operations of a vector-like container, it can
				1097	also perform an 'or' set operation.
				1098	</p>
				1099
				1100	<p>For example:</p>
				1101
				1102	<div class="doc_code">
				1103	<pre>
				1104	enum State {
				1105	None = 0x0,
				1106	FirstCondition = 0x1,
				1107	SecondCondition = 0x2,
				1108	Both = 0x3
				1109	};
				1110
				1111	State get() {
				1112	PackedVector<State, 2> Vec1;
				1113	Vec1.push_back(FirstCondition);
				1114
				1115	PackedVector<State, 2> Vec2;
				1116	Vec2.push_back(SecondCondition);
				1117
				1118	Vec1 \|= Vec2;
				1119	return Vec1[0]; // returns 'Both'.
				1120	}
				1121	</pre>
				1122	</div>
				1123
				1124	</div>
				1125
				1126	<!-- _______________________________________________________________________ -->
				1127	<h4>
Gabor Greif	0186250	2009-02-27 13:28:07 +0000	[diff] [blame]	1128	<a name="dss_ilist_traits">ilist_traits</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1129	</h4>
Gabor Greif	0186250	2009-02-27 13:28:07 +0000	[diff] [blame]	1130
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1131	<div>
Gabor Greif	0186250	2009-02-27 13:28:07 +0000	[diff] [blame]	1132	<p><tt>ilist_traits<T></tt> is <tt>ilist<T></tt>'s customization
				1133	mechanism. <tt>iplist<T></tt> (and consequently <tt>ilist<T></tt>)
				1134	publicly derive from this traits class.</p>
				1135	</div>
				1136
				1137	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1138	<h4>
Gabor Greif	2946d1c	2009-02-27 12:02:19 +0000	[diff] [blame]	1139	<a name="dss_iplist">iplist</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1140	</h4>
Gabor Greif	2946d1c	2009-02-27 12:02:19 +0000	[diff] [blame]	1141
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1142	<div>
Gabor Greif	2946d1c	2009-02-27 12:02:19 +0000	[diff] [blame]	1143	<p><tt>iplist<T></tt> is <tt>ilist<T></tt>'s base and as such
Gabor Greif	0cbcabe	2009-03-12 09:47:03 +0000	[diff] [blame]	1144	supports a slightly narrower interface. Notably, inserters from
				1145	<tt>T&</tt> are absent.</p>
Gabor Greif	0186250	2009-02-27 13:28:07 +0000	[diff] [blame]	1146
				1147	<p><tt>ilist_traits<T></tt> is a public base of this class and can be
				1148	used for a wide variety of customizations.</p>
Gabor Greif	2946d1c	2009-02-27 12:02:19 +0000	[diff] [blame]	1149	</div>
				1150
				1151	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1152	<h4>
Gabor Greif	3899e49	2009-02-27 11:37:41 +0000	[diff] [blame]	1153	<a name="dss_ilist_node">llvm/ADT/ilist_node.h</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1154	</h4>
Gabor Greif	3899e49	2009-02-27 11:37:41 +0000	[diff] [blame]	1155
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1156	<div>
Gabor Greif	3899e49	2009-02-27 11:37:41 +0000	[diff] [blame]	1157	<p><tt>ilist_node<T></tt> implements a the forward and backward links
				1158	that are expected by the <tt>ilist<T></tt> (and analogous containers)
				1159	in the default manner.</p>
				1160
				1161	<p><tt>ilist_node<T></tt>s are meant to be embedded in the node type
Gabor Greif	0cbcabe	2009-03-12 09:47:03 +0000	[diff] [blame]	1162	<tt>T</tt>, usually <tt>T</tt> publicly derives from
				1163	<tt>ilist_node<T></tt>.</p>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1164	</div>
				1165
				1166	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1167	<h4>
Gabor Greif	6a65f42	2009-03-12 10:30:31 +0000	[diff] [blame]	1168	<a name="dss_ilist_sentinel">Sentinels</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1169	</h4>
Gabor Greif	6a65f42	2009-03-12 10:30:31 +0000	[diff] [blame]	1170
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1171	<div>
Dan Gohman	cf0c9bc	2010-02-25 23:51:27 +0000	[diff] [blame]	1172	<p><tt>ilist</tt>s have another specialty that must be considered. To be a good
Gabor Greif	6a65f42	2009-03-12 10:30:31 +0000	[diff] [blame]	1173	citizen in the C++ ecosystem, it needs to support the standard container
				1174	operations, such as <tt>begin</tt> and <tt>end</tt> iterators, etc. Also, the
				1175	<tt>operator--</tt> must work correctly on the <tt>end</tt> iterator in the
				1176	case of non-empty <tt>ilist</tt>s.</p>
				1177
				1178	<p>The only sensible solution to this problem is to allocate a so-called
				1179	<i>sentinel</i> along with the intrusive list, which serves as the <tt>end</tt>
				1180	iterator, providing the back-link to the last element. However conforming to the
				1181	C++ convention it is illegal to <tt>operator++</tt> beyond the sentinel and it
				1182	also must not be dereferenced.</p>
				1183
				1184	<p>These constraints allow for some implementation freedom to the <tt>ilist</tt>
				1185	how to allocate and store the sentinel. The corresponding policy is dictated
				1186	by <tt>ilist_traits<T></tt>. By default a <tt>T</tt> gets heap-allocated
				1187	whenever the need for a sentinel arises.</p>
				1188
				1189	<p>While the default policy is sufficient in most cases, it may break down when
				1190	<tt>T</tt> does not provide a default constructor. Also, in the case of many
				1191	instances of <tt>ilist</tt>s, the memory overhead of the associated sentinels
				1192	is wasted. To alleviate the situation with numerous and voluminous
				1193	<tt>T</tt>-sentinels, sometimes a trick is employed, leading to <i>ghostly
				1194	sentinels</i>.</p>
				1195
				1196	<p>Ghostly sentinels are obtained by specially-crafted <tt>ilist_traits<T></tt>
				1197	which superpose the sentinel with the <tt>ilist</tt> instance in memory. Pointer
				1198	arithmetic is used to obtain the sentinel, which is relative to the
				1199	<tt>ilist</tt>'s <tt>this</tt> pointer. The <tt>ilist</tt> is augmented by an
				1200	extra pointer, which serves as the back-link of the sentinel. This is the only
				1201	field in the ghostly sentinel which can be legally accessed.</p>
				1202	</div>
				1203
				1204	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1205	<h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1206	<a name="dss_other">Other Sequential Container options</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1207	</h4>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1208
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1209	<div>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1210	<p>Other STL containers are available, such as std::string.</p>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1211
				1212	<p>There are also various STL adapter classes such as std::queue,
				1213	std::priority_queue, std::stack, etc. These provide simplified access to an
				1214	underlying container but don't affect the cost of the container itself.</p>
				1215
				1216	</div>
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1217	</div>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1218
				1219	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1220	<h3>
Chris Lattner	7314a20	2011-07-22 20:46:49 +0000	[diff] [blame]	1221	<a name="ds_string">String-like containers</a>
				1222	</h3>
				1223
				1224	<div>
				1225
				1226	<p>
Chris Lattner	3b4f417	2011-07-22 21:34:12 +0000	[diff] [blame]	1227	There are a variety of ways to pass around and use strings in C and C++, and
				1228	LLVM adds a few new options to choose from. Pick the first option on this list
				1229	that will do what you need, they are ordered according to their relative cost.
				1230	</p>
				1231	<p>
				1232	Note that is is generally preferred to <em>not</em> pass strings around as
				1233	"<tt>const char*</tt>"'s. These have a number of problems, including the fact
				1234	that they cannot represent embedded nul ("\0") characters, and do not have a
				1235	length available efficiently. The general replacement for '<tt>const
				1236	char*</tt>' is StringRef.
				1237	</p>
				1238
				1239	<p>For more information on choosing string containers for APIs, please see
				1240	<a href="#string_apis">Passing strings</a>.</p>
				1241
				1242
				1243	<!-- _______________________________________________________________________ -->
				1244	<h4>
				1245	<a name="dss_stringref">llvm/ADT/StringRef.h</a>
				1246	</h4>
				1247
				1248	<div>
				1249	<p>
				1250	The StringRef class is a simple value class that contains a pointer to a
				1251	character and a length, and is quite related to the <a
				1252	href="#dss_arrayref">ArrayRef</a> class (but specialized for arrays of
				1253	characters). Because StringRef carries a length with it, it safely handles
				1254	strings with embedded nul characters in it, getting the length does not require
				1255	a strlen call, and it even has very convenient APIs for slicing and dicing the
				1256	character range that it represents.
				1257	</p>
				1258
				1259	<p>
				1260	StringRef is ideal for passing simple strings around that are known to be live,
				1261	either because they are C string literals, std::string, a C array, or a
				1262	SmallVector. Each of these cases has an efficient implicit conversion to
				1263	StringRef, which doesn't result in a dynamic strlen being executed.
				1264	</p>
				1265
				1266	<p>StringRef has a few major limitations which make more powerful string
				1267	containers useful:</p>
				1268
				1269	<ol>
				1270	<li>You cannot directly convert a StringRef to a 'const char*' because there is
				1271	no way to add a trailing nul (unlike the .c_str() method on various stronger
				1272	classes).</li>
				1273
				1274
				1275	<li>StringRef doesn't own or keep alive the underlying string bytes.
				1276	As such it can easily lead to dangling pointers, and is not suitable for
				1277	embedding in datastructures in most cases (instead, use an std::string or
				1278	something like that).</li>
				1279
				1280	<li>For the same reason, StringRef cannot be used as the return value of a
				1281	method if the method "computes" the result string. Instead, use
				1282	std::string.</li>
				1283
Chris Lattner	ec8f1ea	2011-07-23 17:18:57 +0000	[diff] [blame]	1284	<li>StringRef's do not allow you to mutate the pointed-to string bytes and it
				1285	doesn't allow you to insert or remove bytes from the range. For editing
				1286	operations like this, it interoperates with the <a
				1287	href="#dss_twine">Twine</a> class.</li>
Chris Lattner	3b4f417	2011-07-22 21:34:12 +0000	[diff] [blame]	1288	</ol>
				1289
				1290	<p>Because of its strengths and limitations, it is very common for a function to
				1291	take a StringRef and for a method on an object to return a StringRef that
				1292	points into some string that it owns.</p>
				1293
				1294	</div>
				1295
				1296	<!-- _______________________________________________________________________ -->
				1297	<h4>
				1298	<a name="dss_twine">llvm/ADT/Twine.h</a>
				1299	</h4>
				1300
				1301	<div>
				1302	<p>
				1303	The Twine class is used as an intermediary datatype for APIs that want to take
				1304	a string that can be constructed inline with a series of concatenations.
				1305	Twine works by forming recursive instances of the Twine datatype (a simple
				1306	value object) on the stack as temporary objects, linking them together into a
				1307	tree which is then linearized when the Twine is consumed. Twine is only safe
				1308	to use as the argument to a function, and should always be a const reference,
				1309	e.g.:
				1310	</p>
				1311
				1312	<pre>
				1313	void foo(const Twine &T);
				1314	...
				1315	StringRef X = ...
				1316	unsigned i = ...
				1317	foo(X + "." + Twine(i));
				1318	</pre>
				1319
				1320	<p>This example forms a string like "blarg.42" by concatenating the values
				1321	together, and does not form intermediate strings containing "blarg" or
				1322	"blarg.".
				1323	</p>
				1324
				1325	<p>Because Twine is constructed with temporary objects on the stack, and
				1326	because these instances are destroyed at the end of the current statement,
				1327	it is an inherently dangerous API. For example, this simple variant contains
				1328	undefined behavior and will probably crash:</p>
				1329
				1330	<pre>
				1331	void foo(const Twine &T);
				1332	...
				1333	StringRef X = ...
				1334	unsigned i = ...
				1335	const Twine &Tmp = X + "." + Twine(i);
				1336	foo(Tmp);
				1337	</pre>
				1338
				1339	<p>... because the temporaries are destroyed before the call. That said,
				1340	Twine's are much more efficient than intermediate std::string temporaries, and
				1341	they work really well with StringRef. Just be aware of their limitations.</p>
				1342
				1343	</div>
				1344
				1345
				1346	<!-- _______________________________________________________________________ -->
				1347	<h4>
				1348	<a name="dss_smallstring">llvm/ADT/SmallString.h</a>
				1349	</h4>
				1350
				1351	<div>
				1352
				1353	<p>SmallString is a subclass of <a href="#dss_smallvector">SmallVector</a> that
				1354	adds some convenience APIs like += that takes StringRef's. SmallString avoids
				1355	allocating memory in the case when the preallocated space is enough to hold its
				1356	data, and it calls back to general heap allocation when required. Since it owns
				1357	its data, it is very safe to use and supports full mutation of the string.</p>
				1358
				1359	<p>Like SmallVector's, the big downside to SmallString is their sizeof. While
				1360	they are optimized for small strings, they themselves are not particularly
				1361	small. This means that they work great for temporary scratch buffers on the
				1362	stack, but should not generally be put into the heap: it is very rare to
				1363	see a SmallString as the member of a frequently-allocated heap data structure
				1364	or returned by-value.
Chris Lattner	7314a20	2011-07-22 20:46:49 +0000	[diff] [blame]	1365	</p>
				1366
				1367	</div>
Chris Lattner	3b4f417	2011-07-22 21:34:12 +0000	[diff] [blame]	1368
				1369	<!-- _______________________________________________________________________ -->
				1370	<h4>
				1371	<a name="dss_stdstring">std::string</a>
				1372	</h4>
				1373
				1374	<div>
				1375
				1376	<p>The standard C++ std::string class is a very general class that (like
				1377	SmallString) owns its underlying data. sizeof(std::string) is very reasonable
				1378	so it can be embedded into heap data structures and returned by-value.
				1379	On the other hand, std::string is highly inefficient for inline editing (e.g.
				1380	concatenating a bunch of stuff together) and because it is provided by the
				1381	standard library, its performance characteristics depend a lot of the host
				1382	standard library (e.g. libc++ and MSVC provide a highly optimized string
				1383	class, GCC contains a really slow implementation).
				1384	</p>
				1385
				1386	<p>The major disadvantage of std::string is that almost every operation that
				1387	makes them larger can allocate memory, which is slow. As such, it is better
				1388	to use SmallVector or Twine as a scratch buffer, but then use std::string to
				1389	persist the result.</p>
				1390
				1391
				1392	</div>
				1393
				1394	<!-- end of strings -->
				1395	</div>
				1396
Chris Lattner	7314a20	2011-07-22 20:46:49 +0000	[diff] [blame]	1397
				1398	<!-- ======================================================================= -->
				1399	<h3>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1400	<a name="ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1401	</h3>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1402
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1403	<div>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1404
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1405	<p>Set-like containers are useful when you need to canonicalize multiple values
				1406	into a single representation. There are several different choices for how to do
				1407	this, providing various trade-offs.</p>
				1408
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1409	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1410	<h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1411	<a name="dss_sortedvectorset">A sorted 'vector'</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1412	</h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1413
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1414	<div>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1415
Chris Lattner	3b23a8c	2007-02-03 08:10:45 +0000	[diff] [blame]	1416	<p>If you intend to insert a lot of elements, then do a lot of queries, a
				1417	great approach is to use a vector (or other sequential container) with
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1418	std::sort+std::unique to remove duplicates. This approach works really well if
Chris Lattner	3b23a8c	2007-02-03 08:10:45 +0000	[diff] [blame]	1419	your usage pattern has these two distinct phases (insert then query), and can be
				1420	coupled with a good choice of <a href="#ds_sequential">sequential container</a>.
				1421	</p>
				1422
				1423	<p>
				1424	This combination provides the several nice properties: the result data is
				1425	contiguous in memory (good for cache locality), has few allocations, is easy to
				1426	address (iterators in the final vector are just indices or pointers), and can be
				1427	efficiently queried with a standard binary or radix search.</p>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1428
				1429	</div>
				1430
				1431	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1432	<h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1433	<a name="dss_smallset">"llvm/ADT/SmallSet.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1434	</h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1435
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1436	<div>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1437
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	1438	<p>If you have a set-like data structure that is usually small and whose elements
Chris Lattner	4ddfac1	2007-02-03 07:59:51 +0000	[diff] [blame]	1439	are reasonably small, a <tt>SmallSet<Type, N></tt> is a good choice. This set
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1440	has space for N elements in place (thus, if the set is dynamically smaller than
Chris Lattner	14868db	2007-02-03 08:20:15 +0000	[diff] [blame]	1441	N, no malloc traffic is required) and accesses them with a simple linear search.
				1442	When the set grows beyond 'N' elements, it allocates a more expensive representation that
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1443	guarantees efficient access (for most types, it falls back to std::set, but for
Chris Lattner	14868db	2007-02-03 08:20:15 +0000	[diff] [blame]	1444	pointers it uses something far better, <a
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1445	href="#dss_smallptrset">SmallPtrSet</a>).</p>
				1446
				1447	<p>The magic of this class is that it handles small sets extremely efficiently,
				1448	but gracefully handles extremely large sets without loss of efficiency. The
				1449	drawback is that the interface is quite small: it supports insertion, queries
				1450	and erasing, but does not support iteration.</p>
				1451
				1452	</div>
				1453
				1454	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1455	<h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1456	<a name="dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1457	</h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1458
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1459	<div>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1460
Gabor Greif	4de7368	2010-03-26 19:30:47 +0000	[diff] [blame]	1461	<p>SmallPtrSet has all the advantages of <tt>SmallSet</tt> (and a <tt>SmallSet</tt> of pointers is
				1462	transparently implemented with a <tt>SmallPtrSet</tt>), but also supports iterators. If
Chris Lattner	14868db	2007-02-03 08:20:15 +0000	[diff] [blame]	1463	more than 'N' insertions are performed, a single quadratically
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1464	probed hash table is allocated and grows as needed, providing extremely
				1465	efficient access (constant time insertion/deleting/queries with low constant
				1466	factors) and is very stingy with malloc traffic.</p>
				1467
Gabor Greif	4de7368	2010-03-26 19:30:47 +0000	[diff] [blame]	1468	<p>Note that, unlike <tt>std::set</tt>, the iterators of <tt>SmallPtrSet</tt> are invalidated
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1469	whenever an insertion occurs. Also, the values visited by the iterators are not
				1470	visited in sorted order.</p>
				1471
				1472	</div>
				1473
				1474	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1475	<h4>
Chris Lattner	c28476f	2007-09-30 00:58:59 +0000	[diff] [blame]	1476	<a name="dss_denseset">"llvm/ADT/DenseSet.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1477	</h4>
Chris Lattner	c28476f	2007-09-30 00:58:59 +0000	[diff] [blame]	1478
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1479	<div>
Chris Lattner	c28476f	2007-09-30 00:58:59 +0000	[diff] [blame]	1480
				1481	<p>
				1482	DenseSet is a simple quadratically probed hash table. It excels at supporting
				1483	small values: it uses a single allocation to hold all of the pairs that
				1484	are currently inserted in the set. DenseSet is a great way to unique small
				1485	values that are not simple pointers (use <a
				1486	href="#dss_smallptrset">SmallPtrSet</a> for pointers). Note that DenseSet has
				1487	the same requirements for the value type that <a
				1488	href="#dss_densemap">DenseMap</a> has.
				1489	</p>
				1490
				1491	</div>
				1492
				1493	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1494	<h4>
Jakob Stoklund Olesen	6258862	2012-02-22 00:56:08 +0000	[diff] [blame]	1495	<a name="dss_sparseset">"llvm/ADT/SparseSet.h"</a>
				1496	</h4>
				1497
				1498	<div>
				1499
				1500	<p>SparseSet holds a small number of objects identified by unsigned keys of
				1501	moderate size. It uses a lot of memory, but provides operations that are
				1502	almost as fast as a vector. Typical keys are physical registers, virtual
				1503	registers, or numbered basic blocks.</p>
				1504
				1505	<p>SparseSet is useful for algorithms that need very fast clear/find/insert/erase
				1506	and fast iteration over small sets. It is not intended for building composite
				1507	data structures.</p>
				1508
				1509	</div>
				1510
				1511	<!-- _______________________________________________________________________ -->
				1512	<h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1513	<a name="dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1514	</h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1515
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1516	<div>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1517
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1518	<p>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1519	FoldingSet is an aggregate class that is really good at uniquing
				1520	expensive-to-create or polymorphic objects. It is a combination of a chained
				1521	hash table with intrusive links (uniqued objects are required to inherit from
Chris Lattner	14868db	2007-02-03 08:20:15 +0000	[diff] [blame]	1522	FoldingSetNode) that uses <a href="#dss_smallvector">SmallVector</a> as part of
				1523	its ID process.</p>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1524
Chris Lattner	14868db	2007-02-03 08:20:15 +0000	[diff] [blame]	1525	<p>Consider a case where you want to implement a "getOrCreateFoo" method for
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1526	a complex object (for example, a node in the code generator). The client has a
				1527	description of what it wants to generate (it knows the opcode and all the
				1528	operands), but we don't want to 'new' a node, then try inserting it into a set
Chris Lattner	14868db	2007-02-03 08:20:15 +0000	[diff] [blame]	1529	only to find out it already exists, at which point we would have to delete it
				1530	and return the node that already exists.
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1531	</p>
				1532
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1533	<p>To support this style of client, FoldingSet perform a query with a
				1534	FoldingSetNodeID (which wraps SmallVector) that can be used to describe the
				1535	element that we want to query for. The query either returns the element
				1536	matching the ID or it returns an opaque ID that indicates where insertion should
Chris Lattner	14868db	2007-02-03 08:20:15 +0000	[diff] [blame]	1537	take place. Construction of the ID usually does not require heap traffic.</p>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1538
				1539	<p>Because FoldingSet uses intrusive links, it can support polymorphic objects
				1540	in the set (for example, you can have SDNode instances mixed with LoadSDNodes).
				1541	Because the elements are individually allocated, pointers to the elements are
				1542	stable: inserting or removing elements does not invalidate any pointers to other
				1543	elements.
				1544	</p>
				1545
				1546	</div>
				1547
				1548	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1549	<h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1550	<a name="dss_set"><set></a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1551	</h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1552
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1553	<div>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1554
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1555	<p><tt>std::set</tt> is a reasonable all-around set class, which is decent at
				1556	many things but great at nothing. std::set allocates memory for each element
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1557	inserted (thus it is very malloc intensive) and typically stores three pointers
Chris Lattner	14868db	2007-02-03 08:20:15 +0000	[diff] [blame]	1558	per element in the set (thus adding a large amount of per-element space
				1559	overhead). It offers guaranteed log(n) performance, which is not particularly
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1560	fast from a complexity standpoint (particularly if the elements of the set are
				1561	expensive to compare, like strings), and has extremely high constant factors for
				1562	lookup, insertion and removal.</p>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1563
Chris Lattner	14868db	2007-02-03 08:20:15 +0000	[diff] [blame]	1564	<p>The advantages of std::set are that its iterators are stable (deleting or
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1565	inserting an element from the set does not affect iterators or pointers to other
				1566	elements) and that iteration over the set is guaranteed to be in sorted order.
				1567	If the elements in the set are large, then the relative overhead of the pointers
				1568	and malloc traffic is not a big deal, but if the elements of the set are small,
				1569	std::set is almost never a good choice.</p>
				1570
				1571	</div>
				1572
				1573	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1574	<h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1575	<a name="dss_setvector">"llvm/ADT/SetVector.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1576	</h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1577
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1578	<div>
Chris Lattner	edca3c5	2007-02-04 00:00:26 +0000	[diff] [blame]	1579	<p>LLVM's SetVector<Type> is an adapter class that combines your choice of
				1580	a set-like container along with a <a href="#ds_sequential">Sequential
				1581	Container</a>. The important property
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1582	that this provides is efficient insertion with uniquing (duplicate elements are
				1583	ignored) with iteration support. It implements this by inserting elements into
				1584	both a set-like container and the sequential container, using the set-like
				1585	container for uniquing and the sequential container for iteration.
				1586	</p>
				1587
				1588	<p>The difference between SetVector and other sets is that the order of
				1589	iteration is guaranteed to match the order of insertion into the SetVector.
				1590	This property is really important for things like sets of pointers. Because
				1591	pointer values are non-deterministic (e.g. vary across runs of the program on
Chris Lattner	edca3c5	2007-02-04 00:00:26 +0000	[diff] [blame]	1592	different machines), iterating over the pointers in the set will
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1593	not be in a well-defined order.</p>
				1594
				1595	<p>
				1596	The drawback of SetVector is that it requires twice as much space as a normal
				1597	set and has the sum of constant factors from the set-like container and the
				1598	sequential container that it uses. Use it only if you need to iterate over
				1599	the elements in a deterministic order. SetVector is also expensive to delete
Chris Lattner	edca3c5	2007-02-04 00:00:26 +0000	[diff] [blame]	1600	elements out of (linear time), unless you use it's "pop_back" method, which is
				1601	faster.
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1602	</p>
				1603
Bill Wendling	3478173	2011-10-11 06:33:56 +0000	[diff] [blame]	1604	<p><tt>SetVector</tt> is an adapter class that defaults to
				1605	using <tt>std::vector</tt> and a size 16 <tt>SmallSet</tt> for the underlying
				1606	containers, so it is quite expensive. However,
				1607	<tt>"llvm/ADT/SetVector.h"</tt> also provides a <tt>SmallSetVector</tt>
				1608	class, which defaults to using a <tt>SmallVector</tt> and <tt>SmallSet</tt>
				1609	of a specified size. If you use this, and if your sets are dynamically
				1610	smaller than <tt>N</tt>, you will save a lot of heap traffic.</p>
Chris Lattner	edca3c5	2007-02-04 00:00:26 +0000	[diff] [blame]	1611
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1612	</div>
				1613
				1614	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1615	<h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1616	<a name="dss_uniquevector">"llvm/ADT/UniqueVector.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1617	</h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1618
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1619	<div>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1620
				1621	<p>
				1622	UniqueVector is similar to <a href="#dss_setvector">SetVector</a>, but it
				1623	retains a unique ID for each element inserted into the set. It internally
				1624	contains a map and a vector, and it assigns a unique ID for each value inserted
				1625	into the set.</p>
				1626
				1627	<p>UniqueVector is very expensive: its cost is the sum of the cost of
				1628	maintaining both the map and vector, it has high complexity, high constant
				1629	factors, and produces a lot of malloc traffic. It should be avoided.</p>
				1630
				1631	</div>
				1632
Chris Lattner	2fdd005	2011-11-15 22:40:14 +0000	[diff] [blame]	1633	<!-- _______________________________________________________________________ -->
				1634	<h4>
				1635	<a name="dss_immutableset">"llvm/ADT/ImmutableSet.h"</a>
				1636	</h4>
				1637
				1638	<div>
				1639
				1640	<p>
				1641	ImmutableSet is an immutable (functional) set implementation based on an AVL
				1642	tree.
				1643	Adding or removing elements is done through a Factory object and results in the
				1644	creation of a new ImmutableSet object.
				1645	If an ImmutableSet already exists with the given contents, then the existing one
				1646	is returned; equality is compared with a FoldingSetNodeID.
				1647	The time and space complexity of add or remove operations is logarithmic in the
				1648	size of the original set.
				1649
				1650	<p>
				1651	There is no method for returning an element of the set, you can only check for
				1652	membership.
				1653
				1654	</div>
				1655
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1656
				1657	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1658	<h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1659	<a name="dss_otherset">Other Set-Like Container Options</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1660	</h4>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1661
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1662	<div>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1663
				1664	<p>
				1665	The STL provides several other options, such as std::multiset and the various
Chris Lattner	f1a3082	2009-03-09 05:20:45 +0000	[diff] [blame]	1666	"hash_set" like containers (whether from C++ TR1 or from the SGI library). We
				1667	never use hash_set and unordered_set because they are generally very expensive
				1668	(each insertion requires a malloc) and very non-portable.
				1669	</p>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1670
				1671	<p>std::multiset is useful if you're not interested in elimination of
Chris Lattner	14868db	2007-02-03 08:20:15 +0000	[diff] [blame]	1672	duplicates, but has all the drawbacks of std::set. A sorted vector (where you
				1673	don't delete duplicate entries) or some other approach is almost always
				1674	better.</p>
Chris Lattner	74c4ca1	2007-02-03 07:59:07 +0000	[diff] [blame]	1675
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1676	</div>
				1677
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1678	</div>
				1679
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1680	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1681	<h3>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1682	<a name="ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1683	</h3>
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1684
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1685	<div>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1686	Map-like containers are useful when you want to associate data to a key. As
				1687	usual, there are a lot of different ways to do this. :)
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1688
				1689	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1690	<h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1691	<a name="dss_sortedvectormap">A sorted 'vector'</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1692	</h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1693
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1694	<div>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1695
				1696	<p>
				1697	If your usage pattern follows a strict insert-then-query approach, you can
				1698	trivially use the same approach as <a href="#dss_sortedvectorset">sorted vectors
				1699	for set-like containers</a>. The only difference is that your query function
				1700	(which uses std::lower_bound to get efficient log(n) lookup) should only compare
				1701	the key, not both the key and value. This yields the same advantages as sorted
				1702	vectors for sets.
				1703	</p>
				1704	</div>
				1705
				1706	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1707	<h4>
Chris Lattner	796f9fa	2007-02-08 19:14:21 +0000	[diff] [blame]	1708	<a name="dss_stringmap">"llvm/ADT/StringMap.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1709	</h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1710
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1711	<div>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1712
				1713	<p>
				1714	Strings are commonly used as keys in maps, and they are difficult to support
				1715	efficiently: they are variable length, inefficient to hash and compare when
Chris Lattner	796f9fa	2007-02-08 19:14:21 +0000	[diff] [blame]	1716	long, expensive to copy, etc. StringMap is a specialized container designed to
				1717	cope with these issues. It supports mapping an arbitrary range of bytes to an
				1718	arbitrary other object.</p>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1719
Chris Lattner	796f9fa	2007-02-08 19:14:21 +0000	[diff] [blame]	1720	<p>The StringMap implementation uses a quadratically-probed hash table, where
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1721	the buckets store a pointer to the heap allocated entries (and some other
				1722	stuff). The entries in the map must be heap allocated because the strings are
				1723	variable length. The string data (key) and the element object (value) are
				1724	stored in the same allocation with the string data immediately after the element
				1725	object. This container guarantees the "<tt>(char*)(&Value+1)</tt>" points
				1726	to the key string for a value.</p>
				1727
Chris Lattner	796f9fa	2007-02-08 19:14:21 +0000	[diff] [blame]	1728	<p>The StringMap is very fast for several reasons: quadratic probing is very
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1729	cache efficient for lookups, the hash value of strings in buckets is not
Nick Lewycky	2a80aca	2010-08-01 23:18:45 +0000	[diff] [blame]	1730	recomputed when looking up an element, StringMap rarely has to touch the
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1731	memory for unrelated objects when looking up a value (even when hash collisions
				1732	happen), hash table growth does not recompute the hash values for strings
				1733	already in the table, and each pair in the map is store in a single allocation
				1734	(the string data is stored in the same allocation as the Value of a pair).</p>
				1735
Chris Lattner	796f9fa	2007-02-08 19:14:21 +0000	[diff] [blame]	1736	<p>StringMap also provides query methods that take byte ranges, so it only ever
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1737	copies a string if a value is inserted into the table.</p>
Jim Grosbach	838b7a3	2012-04-18 20:28:55 +0000	[diff] [blame]	1738
				1739	<p>StringMap iteratation order, however, is not guaranteed to be deterministic,
				1740	so any uses which require that should instead use a std::map.</p>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1741	</div>
				1742
				1743	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1744	<h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1745	<a name="dss_indexedmap">"llvm/ADT/IndexedMap.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1746	</h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1747
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1748	<div>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1749	<p>
				1750	IndexedMap is a specialized container for mapping small dense integers (or
				1751	values that can be mapped to small dense integers) to some other type. It is
				1752	internally implemented as a vector with a mapping function that maps the keys to
				1753	the dense integer range.
				1754	</p>
				1755
				1756	<p>
				1757	This is useful for cases like virtual registers in the LLVM code generator: they
				1758	have a dense mapping that is offset by a compile-time constant (the first
				1759	virtual register ID).</p>
				1760
				1761	</div>
				1762
				1763	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1764	<h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1765	<a name="dss_densemap">"llvm/ADT/DenseMap.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1766	</h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1767
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1768	<div>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1769
				1770	<p>
				1771	DenseMap is a simple quadratically probed hash table. It excels at supporting
				1772	small keys and values: it uses a single allocation to hold all of the pairs that
				1773	are currently inserted in the map. DenseMap is a great way to map pointers to
				1774	pointers, or map other small types to each other.
				1775	</p>
				1776
				1777	<p>
				1778	There are several aspects of DenseMap that you should be aware of, however. The
Talin	babd598	2012-01-30 06:55:43 +0000	[diff] [blame]	1779	iterators in a DenseMap are invalidated whenever an insertion occurs, unlike
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1780	map. Also, because DenseMap allocates space for a large number of key/value
Chris Lattner	a4a264d	2007-02-03 20:17:53 +0000	[diff] [blame]	1781	pairs (it starts with 64 by default), it will waste a lot of space if your keys
				1782	or values are large. Finally, you must implement a partial specialization of
Chris Lattner	76c1b97	2007-09-17 18:34:04 +0000	[diff] [blame]	1783	DenseMapInfo for the key that you want, if it isn't already supported. This
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1784	is required to tell DenseMap about two special marker values (which can never be
Chris Lattner	a4a264d	2007-02-03 20:17:53 +0000	[diff] [blame]	1785	inserted into the map) that it needs internally.</p>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1786
Talin	babd598	2012-01-30 06:55:43 +0000	[diff] [blame]	1787	<p>
				1788	DenseMap's find_as() method supports lookup operations using an alternate key
				1789	type. This is useful in cases where the normal key type is expensive to
				1790	construct, but cheap to compare against. The DenseMapInfo is responsible for
				1791	defining the appropriate comparison and hashing methods for each alternate
				1792	key type used.
				1793	</p>
				1794
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1795	</div>
				1796
				1797	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1798	<h4>
Jeffrey Yasskin	71a5c22	2009-10-22 22:11:22 +0000	[diff] [blame]	1799	<a name="dss_valuemap">"llvm/ADT/ValueMap.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1800	</h4>
Jeffrey Yasskin	71a5c22	2009-10-22 22:11:22 +0000	[diff] [blame]	1801
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1802	<div>
Jeffrey Yasskin	71a5c22	2009-10-22 22:11:22 +0000	[diff] [blame]	1803
				1804	<p>
				1805	ValueMap is a wrapper around a <a href="#dss_densemap">DenseMap</a> mapping
				1806	Value*s (or subclasses) to another type. When a Value is deleted or RAUW'ed,
				1807	ValueMap will update itself so the new version of the key is mapped to the same
				1808	value, just as if the key were a WeakVH. You can configure exactly how this
				1809	happens, and what else happens on these two events, by passing
				1810	a <code>Config</code> parameter to the ValueMap template.</p>
				1811
				1812	</div>
				1813
				1814	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1815	<h4>
Jakob Stoklund Olesen	aca0da6	2010-12-14 00:55:51 +0000	[diff] [blame]	1816	<a name="dss_intervalmap">"llvm/ADT/IntervalMap.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1817	</h4>
Jakob Stoklund Olesen	aca0da6	2010-12-14 00:55:51 +0000	[diff] [blame]	1818
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1819	<div>
Jakob Stoklund Olesen	aca0da6	2010-12-14 00:55:51 +0000	[diff] [blame]	1820
				1821	<p> IntervalMap is a compact map for small keys and values. It maps key
				1822	intervals instead of single keys, and it will automatically coalesce adjacent
				1823	intervals. When then map only contains a few intervals, they are stored in the
				1824	map object itself to avoid allocations.</p>
				1825
				1826	<p> The IntervalMap iterators are quite big, so they should not be passed around
				1827	as STL iterators. The heavyweight iterators allow a smaller data structure.</p>
				1828
				1829	</div>
				1830
				1831	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1832	<h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1833	<a name="dss_map"><map></a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1834	</h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1835
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1836	<div>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1837
				1838	<p>
				1839	std::map has similar characteristics to <a href="#dss_set">std::set</a>: it uses
				1840	a single allocation per pair inserted into the map, it offers log(n) lookup with
				1841	an extremely large constant factor, imposes a space penalty of 3 pointers per
				1842	pair in the map, etc.</p>
				1843
				1844	<p>std::map is most useful when your keys or values are very large, if you need
				1845	to iterate over the collection in sorted order, or if you need stable iterators
				1846	into the map (i.e. they don't get invalidated if an insertion or deletion of
				1847	another element takes place).</p>
				1848
				1849	</div>
				1850
				1851	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1852	<h4>
Jakob Stoklund Olesen	4359e5e	2011-04-05 20:56:08 +0000	[diff] [blame]	1853	<a name="dss_inteqclasses">"llvm/ADT/IntEqClasses.h"</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1854	</h4>
Jakob Stoklund Olesen	4359e5e	2011-04-05 20:56:08 +0000	[diff] [blame]	1855
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1856	<div>
Jakob Stoklund Olesen	4359e5e	2011-04-05 20:56:08 +0000	[diff] [blame]	1857
				1858	<p>IntEqClasses provides a compact representation of equivalence classes of
				1859	small integers. Initially, each integer in the range 0..n-1 has its own
				1860	equivalence class. Classes can be joined by passing two class representatives to
				1861	the join(a, b) method. Two integers are in the same class when findLeader()
				1862	returns the same representative.</p>
				1863
				1864	<p>Once all equivalence classes are formed, the map can be compressed so each
				1865	integer 0..n-1 maps to an equivalence class number in the range 0..m-1, where m
				1866	is the total number of equivalence classes. The map must be uncompressed before
				1867	it can be edited again.</p>
				1868
				1869	</div>
				1870
				1871	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1872	<h4>
Chris Lattner	2fdd005	2011-11-15 22:40:14 +0000	[diff] [blame]	1873	<a name="dss_immutablemap">"llvm/ADT/ImmutableMap.h"</a>
				1874	</h4>
				1875
				1876	<div>
				1877
				1878	<p>
				1879	ImmutableMap is an immutable (functional) map implementation based on an AVL
				1880	tree.
				1881	Adding or removing elements is done through a Factory object and results in the
				1882	creation of a new ImmutableMap object.
				1883	If an ImmutableMap already exists with the given key set, then the existing one
				1884	is returned; equality is compared with a FoldingSetNodeID.
				1885	The time and space complexity of add or remove operations is logarithmic in the
				1886	size of the original map.
				1887
				1888	</div>
				1889
				1890	<!-- _______________________________________________________________________ -->
				1891	<h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1892	<a name="dss_othermap">Other Map-Like Container Options</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1893	</h4>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1894
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1895	<div>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1896
				1897	<p>
				1898	The STL provides several other options, such as std::multimap and the various
Chris Lattner	f1a3082	2009-03-09 05:20:45 +0000	[diff] [blame]	1899	"hash_map" like containers (whether from C++ TR1 or from the SGI library). We
				1900	never use hash_set and unordered_set because they are generally very expensive
				1901	(each insertion requires a malloc) and very non-portable.</p>
Chris Lattner	c572243	2007-02-03 19:49:31 +0000	[diff] [blame]	1902
				1903	<p>std::multimap is useful if you want to map a key to multiple values, but has
				1904	all the drawbacks of std::map. A sorted vector or some other approach is almost
				1905	always better.</p>
				1906
Chris Lattner	098129a	2007-02-03 03:04:03 +0000	[diff] [blame]	1907	</div>
				1908
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1909	</div>
				1910
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	1911	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1912	<h3>
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	1913	<a name="ds_bit">Bit storage containers (BitVector, SparseBitVector)</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1914	</h3>
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	1915
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1916	<div>
Chris Lattner	7086ce7	2007-09-25 22:37:50 +0000	[diff] [blame]	1917	<p>Unlike the other containers, there are only two bit storage containers, and
				1918	choosing when to use each is relatively straightforward.</p>
				1919
				1920	<p>One additional option is
				1921	<tt>std::vector<bool></tt>: we discourage its use for two reasons 1) the
				1922	implementation in many common compilers (e.g. commonly available versions of
				1923	GCC) is extremely inefficient and 2) the C++ standards committee is likely to
				1924	deprecate this container and/or change it significantly somehow. In any case,
				1925	please don't use it.</p>
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	1926
				1927	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1928	<h4>
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	1929	<a name="dss_bitvector">BitVector</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1930	</h4>
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	1931
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1932	<div>
Dan Gohman	5f7775c	2010-01-05 18:24:00 +0000	[diff] [blame]	1933	<p> The BitVector container provides a dynamic size set of bits for manipulation.
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	1934	It supports individual bit setting/testing, as well as set operations. The set
				1935	operations take time O(size of bitvector), but operations are performed one word
				1936	at a time, instead of one bit at a time. This makes the BitVector very fast for
				1937	set operations compared to other containers. Use the BitVector when you expect
				1938	the number of set bits to be high (IE a dense set).
				1939	</p>
				1940	</div>
				1941
				1942	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1943	<h4>
Dan Gohman	5f7775c	2010-01-05 18:24:00 +0000	[diff] [blame]	1944	<a name="dss_smallbitvector">SmallBitVector</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1945	</h4>
Dan Gohman	5f7775c	2010-01-05 18:24:00 +0000	[diff] [blame]	1946
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1947	<div>
Dan Gohman	5f7775c	2010-01-05 18:24:00 +0000	[diff] [blame]	1948	<p> The SmallBitVector container provides the same interface as BitVector, but
				1949	it is optimized for the case where only a small number of bits, less than
				1950	25 or so, are needed. It also transparently supports larger bit counts, but
				1951	slightly less efficiently than a plain BitVector, so SmallBitVector should
				1952	only be used when larger counts are rare.
				1953	</p>
				1954
				1955	<p>
				1956	At this time, SmallBitVector does not support set operations (and, or, xor),
				1957	and its operator[] does not provide an assignable lvalue.
				1958	</p>
				1959	</div>
				1960
				1961	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1962	<h4>
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	1963	<a name="dss_sparsebitvector">SparseBitVector</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1964	</h4>
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	1965
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1966	<div>
Daniel Berlin	1939ace	2007-09-24 17:52:25 +0000	[diff] [blame]	1967	<p> The SparseBitVector container is much like BitVector, with one major
				1968	difference: Only the bits that are set, are stored. This makes the
				1969	SparseBitVector much more space efficient than BitVector when the set is sparse,
				1970	as well as making set operations O(number of set bits) instead of O(size of
				1971	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
				1972	(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).
				1973	</p>
				1974	</div>
Chris Lattner	f623a08	2005-10-17 01:36:23 +0000	[diff] [blame]	1975
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1976	</div>
				1977
				1978	</div>
				1979
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	1980	<!-- *********************************************************************** -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1981	<h2>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	1982	<a name="common">Helpful Hints for Common Operations</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1983	</h2>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	1984	<!-- *********************************************************************** -->
				1985
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	1986	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	1987
				1988	<p>This section describes how to perform some very simple transformations of
				1989	LLVM code. This is meant to give examples of common idioms used, showing the
				1990	practical side of LLVM transformations. <p> Because this is a "how-to" section,
				1991	you should also read about the main classes that you will be working with. The
				1992	<a href="#coreclasses">Core LLVM Class Hierarchy Reference</a> contains details
				1993	and descriptions of the main classes that you should know about.</p>
				1994
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	1995	<!-- NOTE: this section should be heavy on example code -->
				1996	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1997	<h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	1998	<a name="inspection">Basic Inspection and Traversal Routines</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	1999	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2000
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2001	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2002
				2003	<p>The LLVM compiler infrastructure have many different data structures that may
				2004	be traversed. Following the example of the C++ standard template library, the
				2005	techniques used to traverse these various data structures are all basically the
				2006	same. For a enumerable sequence of values, the <tt>XXXbegin()</tt> function (or
				2007	method) returns an iterator to the start of the sequence, the <tt>XXXend()</tt>
				2008	function returns an iterator pointing to one past the last valid element of the
				2009	sequence, and there is some <tt>XXXiterator</tt> data type that is common
				2010	between the two operations.</p>
				2011
				2012	<p>Because the pattern for iteration is common across many different aspects of
				2013	the program representation, the standard template library algorithms may be used
				2014	on them, and it is easier to remember how to iterate. First we show a few common
				2015	examples of the data structures that need to be traversed. Other data
				2016	structures are traversed in very similar ways.</p>
				2017
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2018	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2019	<h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2020	<a name="iterate_function">Iterating over the </a><a
				2021	href="#BasicBlock"><tt>BasicBlock</tt></a>s in a <a
				2022	href="#Function"><tt>Function</tt></a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2023	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2024
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2025	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2026
				2027	<p>It's quite common to have a <tt>Function</tt> instance that you'd like to
				2028	transform in some way; in particular, you'd like to manipulate its
				2029	<tt>BasicBlock</tt>s. To facilitate this, you'll need to iterate over all of
				2030	the <tt>BasicBlock</tt>s that constitute the <tt>Function</tt>. The following is
				2031	an example that prints the name of a <tt>BasicBlock</tt> and the number of
				2032	<tt>Instruction</tt>s it contains:</p>
				2033
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2034	<div class="doc_code">
				2035	<pre>
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	2036	// <i>func is a pointer to a Function instance</i>
				2037	for (Function::iterator i = func->begin(), e = func->end(); i != e; ++i)
				2038	// <i>Print out the name of the basic block if it has one, and then the</i>
				2039	// <i>number of instructions that it contains</i>
Chris Lattner	3fee6ed	2009-09-08 05:15:50 +0000	[diff] [blame]	2040	errs() << "Basic block (name=" << i->getName() << ") has "
Bill Wendling	832171c	2006-12-07 20:04:42 +0000	[diff] [blame]	2041	<< i->size() << " instructions.\n";
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2042	</pre>
				2043	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2044
				2045	<p>Note that i can be used as if it were a pointer for the purposes of
Joel Stanley	9b96c44	2002-09-06 21:55:13 +0000	[diff] [blame]	2046	invoking member functions of the <tt>Instruction</tt> class. This is
				2047	because the indirection operator is overloaded for the iterator
Chris Lattner	7496ec5	2003-08-05 22:54:23 +0000	[diff] [blame]	2048	classes. In the above code, the expression <tt>i->size()</tt> is
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2049	exactly equivalent to <tt>(*i).size()</tt> just like you'd expect.</p>
				2050
				2051	</div>
				2052
				2053	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2054	<h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2055	<a name="iterate_basicblock">Iterating over the </a><a
				2056	href="#Instruction"><tt>Instruction</tt></a>s in a <a
				2057	href="#BasicBlock"><tt>BasicBlock</tt></a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2058	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2059
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2060	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2061
				2062	<p>Just like when dealing with <tt>BasicBlock</tt>s in <tt>Function</tt>s, it's
				2063	easy to iterate over the individual instructions that make up
				2064	<tt>BasicBlock</tt>s. Here's a code snippet that prints out each instruction in
				2065	a <tt>BasicBlock</tt>:</p>
				2066
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2067	<div class="doc_code">
Chris Lattner	55c0461	2005-03-06 06:00:13 +0000	[diff] [blame]	2068	<pre>
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	2069	// <i>blk is a pointer to a BasicBlock instance</i>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2070	for (BasicBlock::iterator i = blk->begin(), e = blk->end(); i != e; ++i)
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	2071	// <i>The next statement works since operator<<(ostream&,...)</i>
				2072	// <i>is overloaded for Instruction&</i>
Chris Lattner	3fee6ed	2009-09-08 05:15:50 +0000	[diff] [blame]	2073	errs() << *i << "\n";
Chris Lattner	55c0461	2005-03-06 06:00:13 +0000	[diff] [blame]	2074	</pre>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2075	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2076
				2077	<p>However, this isn't really the best way to print out the contents of a
				2078	<tt>BasicBlock</tt>! Since the ostream operators are overloaded for virtually
				2079	anything you'll care about, you could have just invoked the print routine on the
Chris Lattner	3fee6ed	2009-09-08 05:15:50 +0000	[diff] [blame]	2080	basic block itself: <tt>errs() << *blk << "\n";</tt>.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2081
				2082	</div>
				2083
				2084	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2085	<h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2086	<a name="iterate_institer">Iterating over the </a><a
				2087	href="#Instruction"><tt>Instruction</tt></a>s in a <a
				2088	href="#Function"><tt>Function</tt></a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2089	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2090
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2091	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2092
				2093	<p>If you're finding that you commonly iterate over a <tt>Function</tt>'s
				2094	<tt>BasicBlock</tt>s and then that <tt>BasicBlock</tt>'s <tt>Instruction</tt>s,
				2095	<tt>InstIterator</tt> should be used instead. You'll need to include <a
				2096	href="/doxygen/InstIterator_8h-source.html"><tt>llvm/Support/InstIterator.h</tt></a>,
				2097	and then instantiate <tt>InstIterator</tt>s explicitly in your code. Here's a
Chris Lattner	69bf8a9	2004-05-23 21:06:58 +0000	[diff] [blame]	2098	small example that shows how to dump all instructions in a function to the standard error stream:<p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2099
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2100	<div class="doc_code">
				2101	<pre>
				2102	#include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"
				2103
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	2104	// <i>F is a pointer to a Function instance</i>
Chris Lattner	da021aa	2008-06-04 18:20:42 +0000	[diff] [blame]	2105	for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
Chris Lattner	3fee6ed	2009-09-08 05:15:50 +0000	[diff] [blame]	2106	errs() << *I << "\n";
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2107	</pre>
				2108	</div>
				2109
				2110	<p>Easy, isn't it? You can also use <tt>InstIterator</tt>s to fill a
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	2111	work list with its initial contents. For example, if you wanted to
				2112	initialize a work list to contain all instructions in a <tt>Function</tt>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2113	F, all you would need to do is something like:</p>
				2114
				2115	<div class="doc_code">
				2116	<pre>
				2117	std::set<Instruction*> worklist;
Chris Lattner	da021aa	2008-06-04 18:20:42 +0000	[diff] [blame]	2118	// or better yet, SmallPtrSet<Instruction*, 64> worklist;
				2119
				2120	for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
				2121	worklist.insert(&*I);
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2122	</pre>
				2123	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2124
				2125	<p>The STL set <tt>worklist</tt> would now contain all instructions in the
				2126	<tt>Function</tt> pointed to by F.</p>
				2127
				2128	</div>
				2129
				2130	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2131	<h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2132	<a name="iterate_convert">Turning an iterator into a class pointer (and
				2133	vice-versa)</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2134	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2135
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2136	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2137
				2138	<p>Sometimes, it'll be useful to grab a reference (or pointer) to a class
Joel Stanley	9b96c44	2002-09-06 21:55:13 +0000	[diff] [blame]	2139	instance when all you've got at hand is an iterator. Well, extracting
Chris Lattner	69bf8a9	2004-05-23 21:06:58 +0000	[diff] [blame]	2140	a reference or a pointer from an iterator is very straight-forward.
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	2141	Assuming that <tt>i</tt> is a <tt>BasicBlock::iterator</tt> and <tt>j</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2142	is a <tt>BasicBlock::const_iterator</tt>:</p>
				2143
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2144	<div class="doc_code">
				2145	<pre>
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	2146	Instruction& inst = *i; // <i>Grab reference to instruction reference</i>
				2147	Instruction* pinst = &*i; // <i>Grab pointer to instruction reference</i>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2148	const Instruction& inst = *j;
				2149	</pre>
				2150	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2151
				2152	<p>However, the iterators you'll be working with in the LLVM framework are
				2153	special: they will automatically convert to a ptr-to-instance type whenever they
				2154	need to. Instead of dereferencing the iterator and then taking the address of
				2155	the result, you can simply assign the iterator to the proper pointer type and
				2156	you get the dereference and address-of operation as a result of the assignment
				2157	(behind the scenes, this is a result of overloading casting mechanisms). Thus
				2158	the last line of the last example,</p>
				2159
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2160	<div class="doc_code">
				2161	<pre>
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	2162	Instruction pinst = &i;
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2163	</pre>
				2164	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2165
				2166	<p>is semantically equivalent to</p>
				2167
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2168	<div class="doc_code">
				2169	<pre>
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	2170	Instruction *pinst = i;
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2171	</pre>
				2172	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2173
Chris Lattner	69bf8a9	2004-05-23 21:06:58 +0000	[diff] [blame]	2174	<p>It's also possible to turn a class pointer into the corresponding iterator,
				2175	and this is a constant time operation (very efficient). The following code
				2176	snippet illustrates use of the conversion constructors provided by LLVM
				2177	iterators. By using these, you can explicitly grab the iterator of something
				2178	without actually obtaining it via iteration over some structure:</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2179
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2180	<div class="doc_code">
				2181	<pre>
				2182	void printNextInstruction(Instruction* inst) {
				2183	BasicBlock::iterator it(inst);
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	2184	++it; // <i>After this line, it refers to the instruction after *inst</i>
Chris Lattner	3fee6ed	2009-09-08 05:15:50 +0000	[diff] [blame]	2185	if (it != inst->getParent()->end()) errs() << *it << "\n";
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2186	}
				2187	</pre>
				2188	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2189
Dan Gohman	525bf8e	2010-03-26 19:39:05 +0000	[diff] [blame]	2190	<p>Unfortunately, these implicit conversions come at a cost; they prevent
				2191	these iterators from conforming to standard iterator conventions, and thus
Dan Gohman	0d91c11	2010-03-26 19:51:14 +0000	[diff] [blame]	2192	from being usable with standard algorithms and containers. For example, they
				2193	prevent the following code, where <tt>B</tt> is a <tt>BasicBlock</tt>,
Dan Gohman	525bf8e	2010-03-26 19:39:05 +0000	[diff] [blame]	2194	from compiling:</p>
				2195
				2196	<div class="doc_code">
				2197	<pre>
				2198	llvm::SmallVector<llvm::Instruction *, 16>(B->begin(), B->end());
				2199	</pre>
				2200	</div>
				2201
				2202	<p>Because of this, these implicit conversions may be removed some day,
Dan Gohman	0d91c11	2010-03-26 19:51:14 +0000	[diff] [blame]	2203	and <tt>operator*</tt> changed to return a pointer instead of a reference.</p>
Dan Gohman	525bf8e	2010-03-26 19:39:05 +0000	[diff] [blame]	2204
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2205	</div>
				2206
				2207	<!--_______________________________________________________________________-->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2208	<h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2209	<a name="iterate_complex">Finding call sites: a slightly more complex
				2210	example</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2211	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2212
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2213	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2214
				2215	<p>Say that you're writing a FunctionPass and would like to count all the
				2216	locations in the entire module (that is, across every <tt>Function</tt>) where a
				2217	certain function (i.e., some <tt>Function</tt>*) is already in scope. As you'll
				2218	learn later, you may want to use an <tt>InstVisitor</tt> to accomplish this in a
Chris Lattner	69bf8a9	2004-05-23 21:06:58 +0000	[diff] [blame]	2219	much more straight-forward manner, but this example will allow us to explore how
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	2220	you'd do it if you didn't have <tt>InstVisitor</tt> around. In pseudo-code, this
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2221	is what we want to do:</p>
				2222
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2223	<div class="doc_code">
				2224	<pre>
				2225	initialize callCounter to zero
				2226	for each Function f in the Module
				2227	for each BasicBlock b in f
				2228	for each Instruction i in b
				2229	if (i is a CallInst and calls the given function)
				2230	increment callCounter
				2231	</pre>
				2232	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2233
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2234	<p>And the actual code is (remember, because we're writing a
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2235	<tt>FunctionPass</tt>, our <tt>FunctionPass</tt>-derived class simply has to
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2236	override the <tt>runOnFunction</tt> method):</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2237
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2238	<div class="doc_code">
				2239	<pre>
				2240	Function* targetFunc = ...;
				2241
				2242	class OurFunctionPass : public FunctionPass {
				2243	public:
				2244	OurFunctionPass(): callCounter(0) { }
				2245
				2246	virtual runOnFunction(Function& F) {
				2247	for (Function::iterator b = F.begin(), be = F.end(); b != be; ++b) {
Eric Christopher	203e71d	2008-11-08 08:20:49 +0000	[diff] [blame]	2248	for (BasicBlock::iterator i = b->begin(), ie = b->end(); i != ie; ++i) {
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2249	if (<a href="#CallInst">CallInst</a>* callInst = <a href="#isa">dyn_cast</a><<a
				2250	href="#CallInst">CallInst</a>>(&*i)) {
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	2251	// <i>We know we've encountered a call instruction, so we</i>
				2252	// <i>need to determine if it's a call to the</i>
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	2253	// <i>function pointed to by m_func or not.</i>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2254	if (callInst->getCalledFunction() == targetFunc)
				2255	++callCounter;
				2256	}
				2257	}
				2258	}
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	2259	}
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2260
				2261	private:
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	2262	unsigned callCounter;
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2263	};
				2264	</pre>
				2265	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2266
				2267	</div>
				2268
Brian Gaeke	f1972c6	2003-11-07 19:25:45 +0000	[diff] [blame]	2269	<!--_______________________________________________________________________-->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2270	<h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2271	<a name="calls_and_invokes">Treating calls and invokes the same way</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2272	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2273
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2274	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2275
				2276	<p>You may have noticed that the previous example was a bit oversimplified in
				2277	that it did not deal with call sites generated by 'invoke' instructions. In
				2278	this, and in other situations, you may find that you want to treat
				2279	<tt>CallInst</tt>s and <tt>InvokeInst</tt>s the same way, even though their
				2280	most-specific common base class is <tt>Instruction</tt>, which includes lots of
				2281	less closely-related things. For these cases, LLVM provides a handy wrapper
				2282	class called <a
Reid Spencer	05fe4b0	2006-03-14 05:39:39 +0000	[diff] [blame]	2283	href="http://llvm.org/doxygen/classllvm_1_1CallSite.html"><tt>CallSite</tt></a>.
Chris Lattner	69bf8a9	2004-05-23 21:06:58 +0000	[diff] [blame]	2284	It is essentially a wrapper around an <tt>Instruction</tt> pointer, with some
				2285	methods that provide functionality common to <tt>CallInst</tt>s and
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2286	<tt>InvokeInst</tt>s.</p>
				2287
Chris Lattner	69bf8a9	2004-05-23 21:06:58 +0000	[diff] [blame]	2288	<p>This class has "value semantics": it should be passed by value, not by
				2289	reference and it should not be dynamically allocated or deallocated using
				2290	<tt>operator new</tt> or <tt>operator delete</tt>. It is efficiently copyable,
				2291	assignable and constructable, with costs equivalents to that of a bare pointer.
				2292	If you look at its definition, it has only a single pointer member.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2293
				2294	</div>
				2295
Chris Lattner	1a3105b	2002-09-09 05:49:39 +0000	[diff] [blame]	2296	<!--_______________________________________________________________________-->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2297	<h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2298	<a name="iterate_chains">Iterating over def-use & use-def chains</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2299	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2300
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2301	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2302
				2303	<p>Frequently, we might have an instance of the <a
Chris Lattner	0081517	2007-01-04 22:01:45 +0000	[diff] [blame]	2304	href="/doxygen/classllvm_1_1Value.html">Value Class</a> and we want to
Misha Brukman	384047f	2004-06-03 23:29:12 +0000	[diff] [blame]	2305	determine which <tt>User</tt>s use the <tt>Value</tt>. The list of all
				2306	<tt>User</tt>s of a particular <tt>Value</tt> is called a <i>def-use</i> chain.
				2307	For example, let's say we have a <tt>Function*</tt> named <tt>F</tt> to a
				2308	particular function <tt>foo</tt>. Finding all of the instructions that
				2309	<i>use</i> <tt>foo</tt> is as simple as iterating over the <i>def-use</i> chain
				2310	of <tt>F</tt>:</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2311
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2312	<div class="doc_code">
				2313	<pre>
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	2314	Function *F = ...;
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2315
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	2316	for (Value::use_iterator i = F->use_begin(), e = F->use_end(); i != e; ++i)
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2317	if (Instruction Inst = dyn_cast<Instruction>(i)) {
Chris Lattner	3fee6ed	2009-09-08 05:15:50 +0000	[diff] [blame]	2318	errs() << "F is used in instruction:\n";
				2319	errs() << *Inst << "\n";
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2320	}
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2321	</pre>
Gabor Greif	394fdfb	2010-03-26 19:35:48 +0000	[diff] [blame]	2322	</div>
				2323
Gabor Greif	ce9431953	2010-03-26 19:40:38 +0000	[diff] [blame]	2324	<p>Note that dereferencing a <tt>Value::use_iterator</tt> is not a very cheap
Gabor Greif	4de7368	2010-03-26 19:30:47 +0000	[diff] [blame]	2325	operation. Instead of performing <tt>*i</tt> above several times, consider
Gabor Greif	ce9431953	2010-03-26 19:40:38 +0000	[diff] [blame]	2326	doing it only once in the loop body and reusing its result.</p>
Gabor Greif	4de7368	2010-03-26 19:30:47 +0000	[diff] [blame]	2327
Gabor Greif	6091ff3	2010-03-26 19:04:42 +0000	[diff] [blame]	2328	<p>Alternatively, it's common to have an instance of the <a
Misha Brukman	384047f	2004-06-03 23:29:12 +0000	[diff] [blame]	2329	href="/doxygen/classllvm_1_1User.html">User Class</a> and need to know what
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2330	<tt>Value</tt>s are used by it. The list of all <tt>Value</tt>s used by a
				2331	<tt>User</tt> is known as a <i>use-def</i> chain. Instances of class
				2332	<tt>Instruction</tt> are common <tt>User</tt>s, so we might want to iterate over
				2333	all of the values that a particular instruction uses (that is, the operands of
				2334	the particular <tt>Instruction</tt>):</p>
				2335
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2336	<div class="doc_code">
				2337	<pre>
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	2338	Instruction *pi = ...;
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2339
				2340	for (User::op_iterator i = pi->op_begin(), e = pi->op_end(); i != e; ++i) {
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	2341	Value v = i;
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	2342	// <i>...</i>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2343	}
				2344	</pre>
				2345	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2346
Gabor Greif	4de7368	2010-03-26 19:30:47 +0000	[diff] [blame]	2347	<p>Declaring objects as <tt>const</tt> is an important tool of enforcing
Gabor Greif	ce9431953	2010-03-26 19:40:38 +0000	[diff] [blame]	2348	mutation free algorithms (such as analyses, etc.). For this purpose above
Gabor Greif	4de7368	2010-03-26 19:30:47 +0000	[diff] [blame]	2349	iterators come in constant flavors as <tt>Value::const_use_iterator</tt>
				2350	and <tt>Value::const_op_iterator</tt>. They automatically arise when
				2351	calling <tt>use/op_begin()</tt> on <tt>const Value*</tt>s or
				2352	<tt>const User*</tt>s respectively. Upon dereferencing, they return
Gabor Greif	ce9431953	2010-03-26 19:40:38 +0000	[diff] [blame]	2353	<tt>const Use*</tt>s. Otherwise the above patterns remain unchanged.</p>
				2354
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2355	</div>
				2356
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	2357	<!--_______________________________________________________________________-->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2358	<h4>
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	2359	<a name="iterate_preds">Iterating over predecessors &
				2360	successors of blocks</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2361	</h4>
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	2362
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2363	<div>
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	2364
				2365	<p>Iterating over the predecessors and successors of a block is quite easy
				2366	with the routines defined in <tt>"llvm/Support/CFG.h"</tt>. Just use code like
				2367	this to iterate over all predecessors of BB:</p>
				2368
				2369	<div class="doc_code">
				2370	<pre>
				2371	#include "llvm/Support/CFG.h"
				2372	BasicBlock *BB = ...;
				2373
				2374	for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
				2375	BasicBlock Pred = PI;
				2376	// <i>...</i>
				2377	}
				2378	</pre>
				2379	</div>
				2380
				2381	<p>Similarly, to iterate over successors use
				2382	succ_iterator/succ_begin/succ_end.</p>
				2383
				2384	</div>
				2385
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2386	</div>
Chris Lattner	2e438ca	2008-01-03 16:56:04 +0000	[diff] [blame]	2387
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2388	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2389	<h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2390	<a name="simplechanges">Making simple changes</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2391	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2392
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2393	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2394
				2395	<p>There are some primitive transformation operations present in the LLVM
Joel Stanley	753eb71	2002-09-11 22:32:24 +0000	[diff] [blame]	2396	infrastructure that are worth knowing about. When performing
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	2397	transformations, it's fairly common to manipulate the contents of basic
				2398	blocks. This section describes some of the common methods for doing so
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2399	and gives example code.</p>
				2400
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	2401	<!--_______________________________________________________________________-->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2402	<h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2403	<a name="schanges_creating">Creating and inserting new
				2404	<tt>Instruction</tt>s</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2405	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2406
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2407	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2408
				2409	<p><i>Instantiating Instructions</i></p>
				2410
Chris Lattner	69bf8a9	2004-05-23 21:06:58 +0000	[diff] [blame]	2411	<p>Creation of <tt>Instruction</tt>s is straight-forward: simply call the
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2412	constructor for the kind of instruction to instantiate and provide the necessary
				2413	parameters. For example, an <tt>AllocaInst</tt> only <i>requires</i> a
				2414	(const-ptr-to) <tt>Type</tt>. Thus:</p>
				2415
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2416	<div class="doc_code">
				2417	<pre>
Nick Lewycky	10d64b9	2007-12-03 01:52:52 +0000	[diff] [blame]	2418	AllocaInst* ai = new AllocaInst(Type::Int32Ty);
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2419	</pre>
				2420	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2421
				2422	<p>will create an <tt>AllocaInst</tt> instance that represents the allocation of
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	2423	one integer in the current stack frame, at run time. Each <tt>Instruction</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2424	subclass is likely to have varying default parameters which change the semantics
				2425	of the instruction, so refer to the <a
Misha Brukman	31ca1de	2004-06-03 23:35:54 +0000	[diff] [blame]	2426	href="/doxygen/classllvm_1_1Instruction.html">doxygen documentation for the subclass of
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2427	Instruction</a> that you're interested in instantiating.</p>
				2428
				2429	<p><i>Naming values</i></p>
				2430
				2431	<p>It is very useful to name the values of instructions when you're able to, as
				2432	this facilitates the debugging of your transformations. If you end up looking
				2433	at generated LLVM machine code, you definitely want to have logical names
				2434	associated with the results of instructions! By supplying a value for the
				2435	<tt>Name</tt> (default) parameter of the <tt>Instruction</tt> constructor, you
				2436	associate a logical name with the result of the instruction's execution at
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	2437	run time. For example, say that I'm writing a transformation that dynamically
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2438	allocates space for an integer on the stack, and that integer is going to be
				2439	used as some kind of index by some other code. To accomplish this, I place an
				2440	<tt>AllocaInst</tt> at the first point in the first <tt>BasicBlock</tt> of some
				2441	<tt>Function</tt>, and I'm intending to use it within the same
				2442	<tt>Function</tt>. I might do:</p>
				2443
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2444	<div class="doc_code">
				2445	<pre>
Nick Lewycky	10d64b9	2007-12-03 01:52:52 +0000	[diff] [blame]	2446	AllocaInst* pa = new AllocaInst(Type::Int32Ty, 0, "indexLoc");
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2447	</pre>
				2448	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2449
				2450	<p>where <tt>indexLoc</tt> is now the logical name of the instruction's
Reid Spencer	128a7a7	2007-02-03 21:06:43 +0000	[diff] [blame]	2451	execution value, which is a pointer to an integer on the run time stack.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2452
				2453	<p><i>Inserting instructions</i></p>
				2454
				2455	<p>There are essentially two ways to insert an <tt>Instruction</tt>
				2456	into an existing sequence of instructions that form a <tt>BasicBlock</tt>:</p>
				2457
Joel Stanley	9dd1ad6	2002-09-18 03:17:23 +0000	[diff] [blame]	2458	<ul>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2459	<li>Insertion into an explicit instruction list
				2460
				2461	<p>Given a <tt>BasicBlock* pb</tt>, an <tt>Instruction* pi</tt> within that
				2462	<tt>BasicBlock</tt>, and a newly-created instruction we wish to insert
				2463	before <tt>*pi</tt>, we do the following: </p>
				2464
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2465	<div class="doc_code">
				2466	<pre>
				2467	BasicBlock *pb = ...;
				2468	Instruction *pi = ...;
				2469	Instruction *newInst = new Instruction(...);
				2470
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	2471	pb->getInstList().insert(pi, newInst); // <i>Inserts newInst before pi in pb</i>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2472	</pre>
				2473	</div>
Alkis Evlogimenos	9a5dc4f	2004-05-27 00:57:51 +0000	[diff] [blame]	2474
				2475	<p>Appending to the end of a <tt>BasicBlock</tt> is so common that
				2476	the <tt>Instruction</tt> class and <tt>Instruction</tt>-derived
				2477	classes provide constructors which take a pointer to a
				2478	<tt>BasicBlock</tt> to be appended to. For example code that
				2479	looked like: </p>
				2480
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2481	<div class="doc_code">
				2482	<pre>
				2483	BasicBlock *pb = ...;
				2484	Instruction *newInst = new Instruction(...);
				2485
Bill Wendling	82e2eea	2006-10-11 18:00:22 +0000	[diff] [blame]	2486	pb->getInstList().push_back(newInst); // <i>Appends newInst to pb</i>
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2487	</pre>
				2488	</div>
Alkis Evlogimenos	9a5dc4f	2004-05-27 00:57:51 +0000	[diff] [blame]	2489
				2490	<p>becomes: </p>
				2491
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2492	<div class="doc_code">
				2493	<pre>
				2494	BasicBlock *pb = ...;
				2495	Instruction *newInst = new Instruction(..., pb);
				2496	</pre>
				2497	</div>
Alkis Evlogimenos	9a5dc4f	2004-05-27 00:57:51 +0000	[diff] [blame]	2498
				2499	<p>which is much cleaner, especially if you are creating
				2500	long instruction streams.</p></li>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2501
				2502	<li>Insertion into an implicit instruction list
				2503
				2504	<p><tt>Instruction</tt> instances that are already in <tt>BasicBlock</tt>s
				2505	are implicitly associated with an existing instruction list: the instruction
				2506	list of the enclosing basic block. Thus, we could have accomplished the same
				2507	thing as the above code without being given a <tt>BasicBlock</tt> by doing:
				2508	</p>
				2509
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2510	<div class="doc_code">
				2511	<pre>
				2512	Instruction *pi = ...;
				2513	Instruction *newInst = new Instruction(...);
				2514
				2515	pi->getParent()->getInstList().insert(pi, newInst);
				2516	</pre>
				2517	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2518
				2519	<p>In fact, this sequence of steps occurs so frequently that the
				2520	<tt>Instruction</tt> class and <tt>Instruction</tt>-derived classes provide
				2521	constructors which take (as a default parameter) a pointer to an
				2522	<tt>Instruction</tt> which the newly-created <tt>Instruction</tt> should
				2523	precede. That is, <tt>Instruction</tt> constructors are capable of
				2524	inserting the newly-created instance into the <tt>BasicBlock</tt> of a
				2525	provided instruction, immediately before that instruction. Using an
				2526	<tt>Instruction</tt> constructor with a <tt>insertBefore</tt> (default)
				2527	parameter, the above code becomes:</p>
				2528
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2529	<div class="doc_code">
				2530	<pre>
				2531	Instruction* pi = ...;
				2532	Instruction* newInst = new Instruction(..., pi);
				2533	</pre>
				2534	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2535
				2536	<p>which is much cleaner, especially if you're creating a lot of
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2537	instructions and adding them to <tt>BasicBlock</tt>s.</p></li>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2538	</ul>
				2539
				2540	</div>
				2541
				2542	<!--_______________________________________________________________________-->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2543	<h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2544	<a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2545	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2546
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2547	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2548
				2549	<p>Deleting an instruction from an existing sequence of instructions that form a
Chris Lattner	49e0ccf	2011-03-24 16:13:31 +0000	[diff] [blame]	2550	<a href="#BasicBlock"><tt>BasicBlock</tt></a> is very straight-forward: just
				2551	call the instruction's eraseFromParent() method. For example:</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2552
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2553	<div class="doc_code">
				2554	<pre>
				2555	<a href="#Instruction">Instruction</a> *I = .. ;
Chris Lattner	9f8ec25	2008-02-15 22:57:17 +0000	[diff] [blame]	2556	I->eraseFromParent();
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2557	</pre>
				2558	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2559
Chris Lattner	49e0ccf	2011-03-24 16:13:31 +0000	[diff] [blame]	2560	<p>This unlinks the instruction from its containing basic block and deletes
				2561	it. If you'd just like to unlink the instruction from its containing basic
				2562	block but not delete it, you can use the <tt>removeFromParent()</tt> method.</p>
				2563
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2564	</div>
				2565
				2566	<!--_______________________________________________________________________-->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2567	<h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2568	<a name="schanges_replacing">Replacing an <tt>Instruction</tt> with another
				2569	<tt>Value</tt></a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2570	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2571
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2572	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2573
NAKAMURA Takumi	9c55f59	2012-03-27 11:25:16 +0000	[diff] [blame]	2574	<h5><i>Replacing individual instructions</i></h5>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2575
				2576	<p>Including "<a href="/doxygen/BasicBlockUtils_8h-source.html">llvm/Transforms/Utils/BasicBlockUtils.h</a>"
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	2577	permits use of two very useful replace functions: <tt>ReplaceInstWithValue</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2578	and <tt>ReplaceInstWithInst</tt>.</p>
				2579
NAKAMURA Takumi	06c6d9a	2011-04-18 01:17:51 +0000	[diff] [blame]	2580	<h5><a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a></h5>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2581
NAKAMURA Takumi	9c55f59	2012-03-27 11:25:16 +0000	[diff] [blame]	2582	<div>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	2583	<ul>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2584	<li><tt>ReplaceInstWithValue</tt>
				2585
Nick Lewycky	b6d1f39	2008-09-15 06:31:52 +0000	[diff] [blame]	2586	<p>This function replaces all uses of a given instruction with a value,
				2587	and then removes the original instruction. The following example
				2588	illustrates the replacement of the result of a particular
Chris Lattner	5836082	2005-01-17 00:12:04 +0000	[diff] [blame]	2589	<tt>AllocaInst</tt> that allocates memory for a single integer with a null
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2590	pointer to an integer.</p>
				2591
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2592	<div class="doc_code">
				2593	<pre>
				2594	AllocaInst* instToReplace = ...;
				2595	BasicBlock::iterator ii(instToReplace);
				2596
				2597	ReplaceInstWithValue(instToReplace->getParent()->getInstList(), ii,
Daniel Dunbar	58c2ac0	2008-10-03 22:17:25 +0000	[diff] [blame]	2598	Constant::getNullValue(PointerType::getUnqual(Type::Int32Ty)));
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2599	</pre></div></li>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2600
				2601	<li><tt>ReplaceInstWithInst</tt>
				2602
				2603	<p>This function replaces a particular instruction with another
Nick Lewycky	b6d1f39	2008-09-15 06:31:52 +0000	[diff] [blame]	2604	instruction, inserting the new instruction into the basic block at the
				2605	location where the old instruction was, and replacing any uses of the old
				2606	instruction with the new instruction. The following example illustrates
				2607	the replacement of one <tt>AllocaInst</tt> with another.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2608
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2609	<div class="doc_code">
				2610	<pre>
				2611	AllocaInst* instToReplace = ...;
				2612	BasicBlock::iterator ii(instToReplace);
				2613
				2614	ReplaceInstWithInst(instToReplace->getParent()->getInstList(), ii,
Nick Lewycky	10d64b9	2007-12-03 01:52:52 +0000	[diff] [blame]	2615	new AllocaInst(Type::Int32Ty, 0, "ptrToReplacedInt"));
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	2616	</pre></div></li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	2617	</ul>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2618
NAKAMURA Takumi	9c55f59	2012-03-27 11:25:16 +0000	[diff] [blame]	2619	</div>
				2620
				2621	<h5><i>Replacing multiple uses of <tt>User</tt>s and <tt>Value</tt>s</i></h5>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2622
				2623	<p>You can use <tt>Value::replaceAllUsesWith</tt> and
				2624	<tt>User::replaceUsesOfWith</tt> to change more than one use at a time. See the
Chris Lattner	0081517	2007-01-04 22:01:45 +0000	[diff] [blame]	2625	doxygen documentation for the <a href="/doxygen/classllvm_1_1Value.html">Value Class</a>
Misha Brukman	384047f	2004-06-03 23:29:12 +0000	[diff] [blame]	2626	and <a href="/doxygen/classllvm_1_1User.html">User Class</a>, respectively, for more
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	2627	information.</p>
				2628
				2629	<!-- Value::replaceAllUsesWith User::replaceUsesOfWith Point out:
				2630	include/llvm/Transforms/Utils/ especially BasicBlockUtils.h with:
				2631	ReplaceInstWithValue, ReplaceInstWithInst -->
				2632
				2633	</div>
				2634
Tanya Lattner	b011c66	2007-06-20 18:33:15 +0000	[diff] [blame]	2635	<!--_______________________________________________________________________-->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2636	<h4>
Tanya Lattner	b011c66	2007-06-20 18:33:15 +0000	[diff] [blame]	2637	<a name="schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2638	</h4>
Tanya Lattner	b011c66	2007-06-20 18:33:15 +0000	[diff] [blame]	2639
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2640	<div>
Tanya Lattner	b011c66	2007-06-20 18:33:15 +0000	[diff] [blame]	2641
Tanya Lattner	c5dfcdb	2007-06-20 20:46:37 +0000	[diff] [blame]	2642	<p>Deleting a global variable from a module is just as easy as deleting an
				2643	Instruction. First, you must have a pointer to the global variable that you wish
				2644	to delete. You use this pointer to erase it from its parent, the module.
Tanya Lattner	b011c66	2007-06-20 18:33:15 +0000	[diff] [blame]	2645	For example:</p>
				2646
				2647	<div class="doc_code">
				2648	<pre>
				2649	<a href="#GlobalVariable">GlobalVariable</a> *GV = .. ;
Tanya Lattner	b011c66	2007-06-20 18:33:15 +0000	[diff] [blame]	2650
Tanya Lattner	c5dfcdb	2007-06-20 20:46:37 +0000	[diff] [blame]	2651	GV->eraseFromParent();
Tanya Lattner	b011c66	2007-06-20 18:33:15 +0000	[diff] [blame]	2652	</pre>
				2653	</div>
				2654
				2655	</div>
				2656
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2657	</div>
				2658
Jeffrey Yasskin	714257f	2009-04-30 22:33:41 +0000	[diff] [blame]	2659	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2660	<h3>
Jeffrey Yasskin	714257f	2009-04-30 22:33:41 +0000	[diff] [blame]	2661	<a name="create_types">How to Create Types</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2662	</h3>
Jeffrey Yasskin	714257f	2009-04-30 22:33:41 +0000	[diff] [blame]	2663
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2664	<div>
Jeffrey Yasskin	714257f	2009-04-30 22:33:41 +0000	[diff] [blame]	2665
				2666	<p>In generating IR, you may need some complex types. If you know these types
Misha Brukman	1af789f	2009-05-01 20:40:51 +0000	[diff] [blame]	2667	statically, you can use <tt>TypeBuilder<...>::get()</tt>, defined
Jeffrey Yasskin	714257f	2009-04-30 22:33:41 +0000	[diff] [blame]	2668	in <tt>llvm/Support/TypeBuilder.h</tt>, to retrieve them. <tt>TypeBuilder</tt>
				2669	has two forms depending on whether you're building types for cross-compilation
Misha Brukman	1af789f	2009-05-01 20:40:51 +0000	[diff] [blame]	2670	or native library use. <tt>TypeBuilder<T, true></tt> requires
Jeffrey Yasskin	714257f	2009-04-30 22:33:41 +0000	[diff] [blame]	2671	that <tt>T</tt> be independent of the host environment, meaning that it's built
				2672	out of types from
				2673	the <a href="/doxygen/namespacellvm_1_1types.html"><tt>llvm::types</tt></a>
				2674	namespace and pointers, functions, arrays, etc. built of
Misha Brukman	1af789f	2009-05-01 20:40:51 +0000	[diff] [blame]	2675	those. <tt>TypeBuilder<T, false></tt> additionally allows native C types
Jeffrey Yasskin	714257f	2009-04-30 22:33:41 +0000	[diff] [blame]	2676	whose size may depend on the host compiler. For example,</p>
				2677
				2678	<div class="doc_code">
				2679	<pre>
Misha Brukman	1af789f	2009-05-01 20:40:51 +0000	[diff] [blame]	2680	FunctionType ft = TypeBuilder<types::i<8>(types::i<32>), true>::get();
Jeffrey Yasskin	714257f	2009-04-30 22:33:41 +0000	[diff] [blame]	2681	</pre>
				2682	</div>
				2683
				2684	<p>is easier to read and write than the equivalent</p>
				2685
				2686	<div class="doc_code">
				2687	<pre>
Owen Anderson	5e8c50e	2009-06-16 17:40:28 +0000	[diff] [blame]	2688	std::vector<const Type*> params;
Jeffrey Yasskin	714257f	2009-04-30 22:33:41 +0000	[diff] [blame]	2689	params.push_back(PointerType::getUnqual(Type::Int32Ty));
				2690	FunctionType *ft = FunctionType::get(Type::Int8Ty, params, false);
				2691	</pre>
				2692	</div>
				2693
				2694	<p>See the <a href="/doxygen/TypeBuilder_8h-source.html#l00001">class
				2695	comment</a> for more details.</p>
				2696
				2697	</div>
				2698
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2699	</div>
				2700
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	2701	<!-- *********************************************************************** -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2702	<h2>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2703	<a name="threading">Threads and LLVM</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2704	</h2>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2705	<!-- *********************************************************************** -->
				2706
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2707	<div>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2708	<p>
				2709	This section describes the interaction of the LLVM APIs with multithreading,
				2710	both on the part of client applications, and in the JIT, in the hosted
				2711	application.
				2712	</p>
				2713
				2714	<p>
				2715	Note that LLVM's support for multithreading is still relatively young. Up
				2716	through version 2.5, the execution of threaded hosted applications was
				2717	supported, but not threaded client access to the APIs. While this use case is
				2718	now supported, clients <em>must</em> adhere to the guidelines specified below to
				2719	ensure proper operation in multithreaded mode.
				2720	</p>
				2721
				2722	<p>
				2723	Note that, on Unix-like platforms, LLVM requires the presence of GCC's atomic
				2724	intrinsics in order to support threaded operation. If you need a
				2725	multhreading-capable LLVM on a platform without a suitably modern system
				2726	compiler, consider compiling LLVM and LLVM-GCC in single-threaded mode, and
				2727	using the resultant compiler to build a copy of LLVM with multithreading
				2728	support.
				2729	</p>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2730
				2731	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2732	<h3>
Owen Anderson	1ad70e3	2009-06-16 18:04:19 +0000	[diff] [blame]	2733	<a name="startmultithreaded">Entering and Exiting Multithreaded Mode</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2734	</h3>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2735
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2736	<div>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2737
				2738	<p>
				2739	In order to properly protect its internal data structures while avoiding
Owen Anderson	1ad70e3	2009-06-16 18:04:19 +0000	[diff] [blame]	2740	excessive locking overhead in the single-threaded case, the LLVM must intialize
				2741	certain data structures necessary to provide guards around its internals. To do
				2742	so, the client program must invoke <tt>llvm_start_multithreaded()</tt> before
				2743	making any concurrent LLVM API calls. To subsequently tear down these
				2744	structures, use the <tt>llvm_stop_multithreaded()</tt> call. You can also use
				2745	the <tt>llvm_is_multithreaded()</tt> call to check the status of multithreaded
				2746	mode.
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2747	</p>
				2748
				2749	<p>
Owen Anderson	1ad70e3	2009-06-16 18:04:19 +0000	[diff] [blame]	2750	Note that both of these calls must be made <em>in isolation</em>. That is to
				2751	say that no other LLVM API calls may be executing at any time during the
				2752	execution of <tt>llvm_start_multithreaded()</tt> or <tt>llvm_stop_multithreaded
				2753	</tt>. It's is the client's responsibility to enforce this isolation.
				2754	</p>
				2755
				2756	<p>
				2757	The return value of <tt>llvm_start_multithreaded()</tt> indicates the success or
				2758	failure of the initialization. Failure typically indicates that your copy of
				2759	LLVM was built without multithreading support, typically because GCC atomic
				2760	intrinsics were not found in your system compiler. In this case, the LLVM API
				2761	will not be safe for concurrent calls. However, it <em>will</em> be safe for
Jeffrey Yasskin	01eba39	2010-01-29 19:10:38 +0000	[diff] [blame]	2762	hosting threaded applications in the JIT, though <a href="#jitthreading">care
				2763	must be taken</a> to ensure that side exits and the like do not accidentally
				2764	result in concurrent LLVM API calls.
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2765	</p>
				2766	</div>
				2767
				2768	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2769	<h3>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2770	<a name="shutdown">Ending Execution with <tt>llvm_shutdown()</tt></a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2771	</h3>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2772
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2773	<div>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2774	<p>
				2775	When you are done using the LLVM APIs, you should call <tt>llvm_shutdown()</tt>
Owen Anderson	1ad70e3	2009-06-16 18:04:19 +0000	[diff] [blame]	2776	to deallocate memory used for internal structures. This will also invoke
				2777	<tt>llvm_stop_multithreaded()</tt> if LLVM is operating in multithreaded mode.
				2778	As such, <tt>llvm_shutdown()</tt> requires the same isolation guarantees as
				2779	<tt>llvm_stop_multithreaded()</tt>.
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2780	</p>
				2781
				2782	<p>
				2783	Note that, if you use scope-based shutdown, you can use the
				2784	<tt>llvm_shutdown_obj</tt> class, which calls <tt>llvm_shutdown()</tt> in its
				2785	destructor.
				2786	</div>
				2787
				2788	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2789	<h3>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2790	<a name="managedstatic">Lazy Initialization with <tt>ManagedStatic</tt></a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2791	</h3>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2792
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2793	<div>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2794	<p>
				2795	<tt>ManagedStatic</tt> is a utility class in LLVM used to implement static
				2796	initialization of static resources, such as the global type tables. Before the
				2797	invocation of <tt>llvm_shutdown()</tt>, it implements a simple lazy
				2798	initialization scheme. Once <tt>llvm_start_multithreaded()</tt> returns,
				2799	however, it uses double-checked locking to implement thread-safe lazy
				2800	initialization.
				2801	</p>
				2802
				2803	<p>
				2804	Note that, because no other threads are allowed to issue LLVM API calls before
				2805	<tt>llvm_start_multithreaded()</tt> returns, it is possible to have
				2806	<tt>ManagedStatic</tt>s of <tt>llvm::sys::Mutex</tt>s.
				2807	</p>
Owen Anderson	1ad70e3	2009-06-16 18:04:19 +0000	[diff] [blame]	2808
				2809	<p>
				2810	The <tt>llvm_acquire_global_lock()</tt> and <tt>llvm_release_global_lock</tt>
				2811	APIs provide access to the global lock used to implement the double-checked
				2812	locking for lazy initialization. These should only be used internally to LLVM,
				2813	and only if you know what you're doing!
				2814	</p>
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2815	</div>
				2816
Owen Anderson	e0c951a	2009-08-19 17:58:52 +0000	[diff] [blame]	2817	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2818	<h3>
Owen Anderson	e0c951a	2009-08-19 17:58:52 +0000	[diff] [blame]	2819	<a name="llvmcontext">Achieving Isolation with <tt>LLVMContext</tt></a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2820	</h3>
Owen Anderson	e0c951a	2009-08-19 17:58:52 +0000	[diff] [blame]	2821
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2822	<div>
Owen Anderson	e0c951a	2009-08-19 17:58:52 +0000	[diff] [blame]	2823	<p>
				2824	<tt>LLVMContext</tt> is an opaque class in the LLVM API which clients can use
				2825	to operate multiple, isolated instances of LLVM concurrently within the same
				2826	address space. For instance, in a hypothetical compile-server, the compilation
				2827	of an individual translation unit is conceptually independent from all the
				2828	others, and it would be desirable to be able to compile incoming translation
				2829	units concurrently on independent server threads. Fortunately,
				2830	<tt>LLVMContext</tt> exists to enable just this kind of scenario!
				2831	</p>
				2832
				2833	<p>
				2834	Conceptually, <tt>LLVMContext</tt> provides isolation. Every LLVM entity
				2835	(<tt>Module</tt>s, <tt>Value</tt>s, <tt>Type</tt>s, <tt>Constant</tt>s, etc.)
Chris Lattner	38eee3c	2009-08-20 03:10:14 +0000	[diff] [blame]	2836	in LLVM's in-memory IR belongs to an <tt>LLVMContext</tt>. Entities in
Owen Anderson	e0c951a	2009-08-19 17:58:52 +0000	[diff] [blame]	2837	different contexts <em>cannot</em> interact with each other: <tt>Module</tt>s in
				2838	different contexts cannot be linked together, <tt>Function</tt>s cannot be added
				2839	to <tt>Module</tt>s in different contexts, etc. What this means is that is is
				2840	safe to compile on multiple threads simultaneously, as long as no two threads
				2841	operate on entities within the same context.
				2842	</p>
				2843
				2844	<p>
				2845	In practice, very few places in the API require the explicit specification of a
				2846	<tt>LLVMContext</tt>, other than the <tt>Type</tt> creation/lookup APIs.
				2847	Because every <tt>Type</tt> carries a reference to its owning context, most
				2848	other entities can determine what context they belong to by looking at their
				2849	own <tt>Type</tt>. If you are adding new entities to LLVM IR, please try to
				2850	maintain this interface design.
				2851	</p>
				2852
				2853	<p>
				2854	For clients that do <em>not</em> require the benefits of isolation, LLVM
				2855	provides a convenience API <tt>getGlobalContext()</tt>. This returns a global,
				2856	lazily initialized <tt>LLVMContext</tt> that may be used in situations where
				2857	isolation is not a concern.
				2858	</p>
				2859	</div>
				2860
Jeffrey Yasskin	01eba39	2010-01-29 19:10:38 +0000	[diff] [blame]	2861	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2862	<h3>
Jeffrey Yasskin	01eba39	2010-01-29 19:10:38 +0000	[diff] [blame]	2863	<a name="jitthreading">Threads and the JIT</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2864	</h3>
Jeffrey Yasskin	01eba39	2010-01-29 19:10:38 +0000	[diff] [blame]	2865
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2866	<div>
Jeffrey Yasskin	01eba39	2010-01-29 19:10:38 +0000	[diff] [blame]	2867	<p>
				2868	LLVM's "eager" JIT compiler is safe to use in threaded programs. Multiple
				2869	threads can call <tt>ExecutionEngine::getPointerToFunction()</tt> or
				2870	<tt>ExecutionEngine::runFunction()</tt> concurrently, and multiple threads can
				2871	run code output by the JIT concurrently. The user must still ensure that only
				2872	one thread accesses IR in a given <tt>LLVMContext</tt> while another thread
				2873	might be modifying it. One way to do that is to always hold the JIT lock while
				2874	accessing IR outside the JIT (the JIT <em>modifies</em> the IR by adding
				2875	<tt>CallbackVH</tt>s). Another way is to only
				2876	call <tt>getPointerToFunction()</tt> from the <tt>LLVMContext</tt>'s thread.
				2877	</p>
				2878
				2879	<p>When the JIT is configured to compile lazily (using
				2880	<tt>ExecutionEngine::DisableLazyCompilation(false)</tt>), there is currently a
				2881	<a href="http://llvm.org/bugs/show_bug.cgi?id=5184">race condition</a> in
				2882	updating call sites after a function is lazily-jitted. It's still possible to
				2883	use the lazy JIT in a threaded program if you ensure that only one thread at a
				2884	time can call any particular lazy stub and that the JIT lock guards any IR
				2885	access, but we suggest using only the eager JIT in threaded programs.
				2886	</p>
				2887	</div>
				2888
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2889	</div>
				2890
Owen Anderson	8bc1b3b	2009-06-16 01:17:16 +0000	[diff] [blame]	2891	<!-- *********************************************************************** -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2892	<h2>
Chris Lattner	d9d6e10	2005-04-23 16:10:52 +0000	[diff] [blame]	2893	<a name="advanced">Advanced Topics</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2894	</h2>
Chris Lattner	d9d6e10	2005-04-23 16:10:52 +0000	[diff] [blame]	2895	<!-- *********************************************************************** -->
				2896
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2897	<div>
Chris Lattner	f1b200b	2005-04-23 17:27:36 +0000	[diff] [blame]	2898	<p>
				2899	This section describes some of the advanced or obscure API's that most clients
				2900	do not need to be aware of. These API's tend manage the inner workings of the
				2901	LLVM system, and only need to be accessed in unusual circumstances.
				2902	</p>
Chris Lattner	d9d6e10	2005-04-23 16:10:52 +0000	[diff] [blame]	2903
Chris Lattner	1afcace	2011-07-09 17:41:24 +0000	[diff] [blame]	2904
Chris Lattner	f1b200b	2005-04-23 17:27:36 +0000	[diff] [blame]	2905	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2906	<h3>
Chris Lattner	1afcace	2011-07-09 17:41:24 +0000	[diff] [blame]	2907	<a name="SymbolTable">The <tt>ValueSymbolTable</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2908	</h3>
Chris Lattner	f1b200b	2005-04-23 17:27:36 +0000	[diff] [blame]	2909
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2910	<div>
Chris Lattner	263a98e	2007-02-16 04:37:31 +0000	[diff] [blame]	2911	<p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1ValueSymbolTable.html">
				2912	ValueSymbolTable</a></tt> class provides a symbol table that the <a
Chris Lattner	d9d6e10	2005-04-23 16:10:52 +0000	[diff] [blame]	2913	href="#Function"><tt>Function</tt></a> and <a href="#Module">
Chris Lattner	263a98e	2007-02-16 04:37:31 +0000	[diff] [blame]	2914	<tt>Module</tt></a> classes use for naming value definitions. The symbol table
				2915	can provide a name for any <a href="#Value"><tt>Value</tt></a>.
Chris Lattner	1afcace	2011-07-09 17:41:24 +0000	[diff] [blame]	2916	</p>
Chris Lattner	d9d6e10	2005-04-23 16:10:52 +0000	[diff] [blame]	2917
Reid Spencer	a636224	2007-01-07 00:41:39 +0000	[diff] [blame]	2918	<p>Note that the <tt>SymbolTable</tt> class should not be directly accessed
				2919	by most clients. It should only be used when iteration over the symbol table
				2920	names themselves are required, which is very special purpose. Note that not
				2921	all LLVM
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2922	<tt><a href="#Value">Value</a></tt>s have names, and those without names (i.e. they have
Chris Lattner	d9d6e10	2005-04-23 16:10:52 +0000	[diff] [blame]	2923	an empty name) do not exist in the symbol table.
				2924	</p>
				2925
Chris Lattner	1afcace	2011-07-09 17:41:24 +0000	[diff] [blame]	2926	<p>Symbol tables support iteration over the values in the symbol
Chris Lattner	263a98e	2007-02-16 04:37:31 +0000	[diff] [blame]	2927	table with <tt>begin/end/iterator</tt> and supports querying to see if a
				2928	specific name is in the symbol table (with <tt>lookup</tt>). The
				2929	<tt>ValueSymbolTable</tt> class exposes no public mutator methods, instead,
				2930	simply call <tt>setName</tt> on a value, which will autoinsert it into the
Chris Lattner	1afcace	2011-07-09 17:41:24 +0000	[diff] [blame]	2931	appropriate symbol table.</p>
Chris Lattner	d9d6e10	2005-04-23 16:10:52 +0000	[diff] [blame]	2932
Chris Lattner	d9d6e10	2005-04-23 16:10:52 +0000	[diff] [blame]	2933	</div>
				2934
				2935
				2936
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2937	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2938	<h3>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2939	<a name="UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2940	</h3>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2941
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2942	<div>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2943	<p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1User.html">
Gabor Greif	fd095b6	2009-01-05 16:05:32 +0000	[diff] [blame]	2944	User</a></tt> class provides a basis for expressing the ownership of <tt>User</tt>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2945	towards other <tt><a href="http://llvm.org/doxygen/classllvm_1_1Value.html">
				2946	Value</a></tt>s. The <tt><a href="http://llvm.org/doxygen/classllvm_1_1Use.html">
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	2947	Use</a></tt> helper class is employed to do the bookkeeping and to facilitate <i>O(1)</i>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2948	addition and removal.</p>
				2949
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	2950	<!-- ______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	2951	<h4>
				2952	<a name="Use2User">
				2953	Interaction and relationship between <tt>User</tt> and <tt>Use</tt> objects
				2954	</a>
				2955	</h4>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2956
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	2957	<div>
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	2958	<p>
				2959	A subclass of <tt>User</tt> can choose between incorporating its <tt>Use</tt> objects
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2960	or refer to them out-of-line by means of a pointer. A mixed variant
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	2961	(some <tt>Use</tt>s inline others hung off) is impractical and breaks the invariant
				2962	that the <tt>Use</tt> objects belonging to the same <tt>User</tt> form a contiguous array.
				2963	</p>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2964
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	2965	<p>
				2966	We have 2 different layouts in the <tt>User</tt> (sub)classes:
				2967	<ul>
				2968	<li><p>Layout a)
				2969	The <tt>Use</tt> object(s) are inside (resp. at fixed offset) of the <tt>User</tt>
				2970	object and there are a fixed number of them.</p>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2971
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	2972	<li><p>Layout b)
				2973	The <tt>Use</tt> object(s) are referenced by a pointer to an
				2974	array from the <tt>User</tt> object and there may be a variable
				2975	number of them.</p>
				2976	</ul>
				2977	<p>
Gabor Greif	d41720a	2008-06-25 00:10:22 +0000	[diff] [blame]	2978	As of v2.4 each layout still possesses a direct pointer to the
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	2979	start of the array of <tt>Use</tt>s. Though not mandatory for layout a),
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2980	we stick to this redundancy for the sake of simplicity.
Gabor Greif	d41720a	2008-06-25 00:10:22 +0000	[diff] [blame]	2981	The <tt>User</tt> object also stores the number of <tt>Use</tt> objects it
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2982	has. (Theoretically this information can also be calculated
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	2983	given the scheme presented below.)</p>
				2984	<p>
				2985	Special forms of allocation operators (<tt>operator new</tt>)
Gabor Greif	d41720a	2008-06-25 00:10:22 +0000	[diff] [blame]	2986	enforce the following memory layouts:</p>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2987
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	2988	<ul>
Gabor Greif	d41720a	2008-06-25 00:10:22 +0000	[diff] [blame]	2989	<li><p>Layout a) is modelled by prepending the <tt>User</tt> object by the <tt>Use[]</tt> array.</p>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2990
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	2991	<pre>
				2992	...---.---.---.---.-------...
				2993	\| P \| P \| P \| P \| User
				2994	'''---'---'---'---'-------'''
				2995	</pre>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	2996
Gabor Greif	d41720a	2008-06-25 00:10:22 +0000	[diff] [blame]	2997	<li><p>Layout b) is modelled by pointing at the <tt>Use[]</tt> array.</p>
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	2998	<pre>
				2999	.-------...
				3000	\| User
				3001	'-------'''
				3002	\|
				3003	v
				3004	.---.---.---.---...
				3005	\| P \| P \| P \| P \|
				3006	'---'---'---'---'''
				3007	</pre>
				3008	</ul>
				3009	<i>(In the above figures '<tt>P</tt>' stands for the <tt>Use**</tt> that
				3010	is stored in each <tt>Use</tt> object in the member <tt>Use::Prev</tt>)</i>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3011
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3012	</div>
				3013
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3014	<!-- ______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3015	<h4>
Gabor Greif	fd095b6	2009-01-05 16:05:32 +0000	[diff] [blame]	3016	<a name="Waymarking">The waymarking algorithm</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3017	</h4>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3018
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3019	<div>
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3020	<p>
Gabor Greif	d41720a	2008-06-25 00:10:22 +0000	[diff] [blame]	3021	Since the <tt>Use</tt> objects are deprived of the direct (back)pointer to
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3022	their <tt>User</tt> objects, there must be a fast and exact method to
				3023	recover it. This is accomplished by the following scheme:</p>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3024
Gabor Greif	d41720a	2008-06-25 00:10:22 +0000	[diff] [blame]	3025	A bit-encoding in the 2 LSBits (least significant bits) of the <tt>Use::Prev</tt> allows to find the
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3026	start of the <tt>User</tt> object:
				3027	<ul>
				3028	<li><tt>00</tt> —> binary digit 0</li>
				3029	<li><tt>01</tt> —> binary digit 1</li>
				3030	<li><tt>10</tt> —> stop and calculate (<tt>s</tt>)</li>
				3031	<li><tt>11</tt> —> full stop (<tt>S</tt>)</li>
				3032	</ul>
				3033	<p>
				3034	Given a <tt>Use*</tt>, all we have to do is to walk till we get
				3035	a stop and we either have a <tt>User</tt> immediately behind or
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3036	we have to walk to the next stop picking up digits
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3037	and calculating the offset:</p>
				3038	<pre>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3039	.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.----------------
				3040	\| 1 \| s \| 1 \| 0 \| 1 \| 0 \| s \| 1 \| 1 \| 0 \| s \| 1 \| 1 \| s \| 1 \| S \| User (or User*)
				3041	'---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'----------------
				3042	\|+15 \|+10 \|+6 \|+3 \|+1
				3043	\| \| \| \| \|__>
				3044	\| \| \| \|__________>
				3045	\| \| \|______________________>
				3046	\| \|______________________________________>
				3047	\|__________________________________________________________>
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3048	</pre>
				3049	<p>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3050	Only the significant number of bits need to be stored between the
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3051	stops, so that the <i>worst case is 20 memory accesses</i> when there are
				3052	1000 <tt>Use</tt> objects associated with a <tt>User</tt>.</p>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3053
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3054	</div>
				3055
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3056	<!-- ______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3057	<h4>
Gabor Greif	fd095b6	2009-01-05 16:05:32 +0000	[diff] [blame]	3058	<a name="ReferenceImpl">Reference implementation</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3059	</h4>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3060
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3061	<div>
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3062	<p>
				3063	The following literate Haskell fragment demonstrates the concept:</p>
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3064
				3065	<div class="doc_code">
				3066	<pre>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3067	> import Test.QuickCheck
				3068	>
				3069	> digits :: Int -> [Char] -> [Char]
				3070	> digits 0 acc = '0' : acc
				3071	> digits 1 acc = '1' : acc
				3072	> digits n acc = digits (n `div` 2) $ digits (n `mod` 2) acc
				3073	>
				3074	> dist :: Int -> [Char] -> [Char]
				3075	> dist 0 [] = ['S']
				3076	> dist 0 acc = acc
				3077	> dist 1 acc = let r = dist 0 acc in 's' : digits (length r) r
				3078	> dist n acc = dist (n - 1) $ dist 1 acc
				3079	>
				3080	> takeLast n ss = reverse $ take n $ reverse ss
				3081	>
				3082	> test = takeLast 40 $ dist 20 []
				3083	>
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3084	</pre>
				3085	</div>
				3086	<p>
				3087	Printing <test> gives: <tt>"1s100000s11010s10100s1111s1010s110s11s1S"</tt></p>
				3088	<p>
				3089	The reverse algorithm computes the length of the string just by examining
				3090	a certain prefix:</p>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3091
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3092	<div class="doc_code">
				3093	<pre>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3094	> pref :: [Char] -> Int
				3095	> pref "S" = 1
				3096	> pref ('s':'1':rest) = decode 2 1 rest
				3097	> pref (_:rest) = 1 + pref rest
				3098	>
				3099	> decode walk acc ('0':rest) = decode (walk + 1) (acc * 2) rest
				3100	> decode walk acc ('1':rest) = decode (walk + 1) (acc * 2 + 1) rest
				3101	> decode walk acc _ = walk + acc
				3102	>
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3103	</pre>
				3104	</div>
				3105	<p>
				3106	Now, as expected, printing <pref test> gives <tt>40</tt>.</p>
				3107	<p>
				3108	We can <i>quickCheck</i> this with following property:</p>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3109
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3110	<div class="doc_code">
				3111	<pre>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3112	> testcase = dist 2000 []
				3113	> testcaseLength = length testcase
				3114	>
				3115	> identityProp n = n > 0 && n <= testcaseLength ==> length arr == pref arr
				3116	> where arr = takeLast n testcase
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3117	>
				3118	</pre>
				3119	</div>
				3120	<p>
				3121	As expected <quickCheck identityProp> gives:</p>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3122
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3123	<pre>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3124	*Main> quickCheck identityProp
				3125	OK, passed 100 tests.
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3126	</pre>
				3127	<p>
				3128	Let's be a bit more exhaustive:</p>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3129
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3130	<div class="doc_code">
				3131	<pre>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3132	>
				3133	> deepCheck p = check (defaultConfig { configMaxTest = 500 }) p
				3134	>
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3135	</pre>
				3136	</div>
				3137	<p>
				3138	And here is the result of <deepCheck identityProp>:</p>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3139
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3140	<pre>
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3141	*Main> deepCheck identityProp
				3142	OK, passed 500 tests.
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3143	</pre>
				3144
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3145	</div>
				3146
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3147	<!-- ______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3148	<h4>
Gabor Greif	fd095b6	2009-01-05 16:05:32 +0000	[diff] [blame]	3149	<a name="Tagging">Tagging considerations</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3150	</h4>
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3151
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3152	<div>
				3153
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3154	<p>
				3155	To maintain the invariant that the 2 LSBits of each <tt>Use**</tt> in <tt>Use</tt>
				3156	never change after being set up, setters of <tt>Use::Prev</tt> must re-tag the
				3157	new <tt>Use**</tt> on every modification. Accordingly getters must strip the
				3158	tag bits.</p>
				3159	<p>
Gabor Greif	d41720a	2008-06-25 00:10:22 +0000	[diff] [blame]	3160	For layout b) instead of the <tt>User</tt> we find a pointer (<tt>User*</tt> with LSBit set).
				3161	Following this pointer brings us to the <tt>User</tt>. A portable trick ensures
				3162	that the first bytes of <tt>User</tt> (if interpreted as a pointer) never has
Gabor Greif	fd095b6	2009-01-05 16:05:32 +0000	[diff] [blame]	3163	the LSBit set. (Portability is relying on the fact that all known compilers place the
				3164	<tt>vptr</tt> in the first word of the instances.)</p>
Gabor Greif	dfed118	2008-06-18 13:44:57 +0000	[diff] [blame]	3165
Gabor Greif	e98fc27	2008-06-16 21:06:12 +0000	[diff] [blame]	3166	</div>
				3167
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3168	</div>
				3169
				3170	</div>
				3171
				3172	<!-- *********************************************************************** -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3173	<h2>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3174	<a name="coreclasses">The Core LLVM Class Hierarchy Reference </a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3175	</h2>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3176	<!-- *********************************************************************** -->
				3177
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3178	<div>
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3179	<p><tt>#include "<a href="/doxygen/Type_8h-source.html">llvm/Type.h</a>"</tt>
				3180	<br>doxygen info: <a href="/doxygen/classllvm_1_1Type.html">Type Class</a></p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3181
				3182	<p>The Core LLVM classes are the primary means of representing the program
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3183	being inspected or transformed. The core LLVM classes are defined in
				3184	header files in the <tt>include/llvm/</tt> directory, and implemented in
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3185	the <tt>lib/VMCore</tt> directory.</p>
				3186
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3187	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3188	<h3>
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3189	<a name="Type">The <tt>Type</tt> class and Derived Types</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3190	</h3>
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3191
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3192	<div>
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3193
				3194	<p><tt>Type</tt> is a superclass of all type classes. Every <tt>Value</tt> has
				3195	a <tt>Type</tt>. <tt>Type</tt> cannot be instantiated directly but only
				3196	through its subclasses. Certain primitive types (<tt>VoidType</tt>,
				3197	<tt>LabelType</tt>, <tt>FloatType</tt> and <tt>DoubleType</tt>) have hidden
				3198	subclasses. They are hidden because they offer no useful functionality beyond
				3199	what the <tt>Type</tt> class offers except to distinguish themselves from
				3200	other subclasses of <tt>Type</tt>.</p>
				3201	<p>All other types are subclasses of <tt>DerivedType</tt>. Types can be
				3202	named, but this is not a requirement. There exists exactly
				3203	one instance of a given shape at any one time. This allows type equality to
				3204	be performed with address equality of the Type Instance. That is, given two
				3205	<tt>Type*</tt> values, the types are identical if the pointers are identical.
				3206	</p>
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3207
				3208	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3209	<h4>
Gabor Greif	fd095b6	2009-01-05 16:05:32 +0000	[diff] [blame]	3210	<a name="m_Type">Important Public Methods</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3211	</h4>
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3212
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3213	<div>
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3214
				3215	<ul>
Duncan Sands	b0bc6c3	2010-02-15 16:12:20 +0000	[diff] [blame]	3216	<li><tt>bool isIntegerTy() const</tt>: Returns true for any integer type.</li>
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3217
Duncan Sands	b0bc6c3	2010-02-15 16:12:20 +0000	[diff] [blame]	3218	<li><tt>bool isFloatingPointTy()</tt>: Return true if this is one of the five
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3219	floating point types.</li>
				3220
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3221	<li><tt>bool isSized()</tt>: Return true if the type has known size. Things
				3222	that don't have a size are abstract types, labels and void.</li>
				3223
				3224	</ul>
				3225	</div>
				3226
				3227	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3228	<h4>
Gabor Greif	fd095b6	2009-01-05 16:05:32 +0000	[diff] [blame]	3229	<a name="derivedtypes">Important Derived Types</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3230	</h4>
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3231	<div>
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3232	<dl>
				3233	<dt><tt>IntegerType</tt></dt>
				3234	<dd>Subclass of DerivedType that represents integer types of any bit width.
				3235	Any bit width between <tt>IntegerType::MIN_INT_BITS</tt> (1) and
				3236	<tt>IntegerType::MAX_INT_BITS</tt> (~8 million) can be represented.
				3237	<ul>
				3238	<li><tt>static const IntegerType* get(unsigned NumBits)</tt>: get an integer
				3239	type of a specific bit width.</li>
				3240	<li><tt>unsigned getBitWidth() const</tt>: Get the bit width of an integer
				3241	type.</li>
				3242	</ul>
				3243	</dd>
				3244	<dt><tt>SequentialType</tt></dt>
Tobias Grosser	d475c10	2011-07-12 11:37:02 +0000	[diff] [blame]	3245	<dd>This is subclassed by ArrayType, PointerType and VectorType.
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3246	<ul>
				3247	<li><tt>const Type * getElementType() const</tt>: Returns the type of each
				3248	of the elements in the sequential type. </li>
				3249	</ul>
				3250	</dd>
				3251	<dt><tt>ArrayType</tt></dt>
				3252	<dd>This is a subclass of SequentialType and defines the interface for array
				3253	types.
				3254	<ul>
				3255	<li><tt>unsigned getNumElements() const</tt>: Returns the number of
				3256	elements in the array. </li>
				3257	</ul>
				3258	</dd>
				3259	<dt><tt>PointerType</tt></dt>
Chris Lattner	302da1e	2007-02-03 03:05:57 +0000	[diff] [blame]	3260	<dd>Subclass of SequentialType for pointer types.</dd>
Reid Spencer	9d6565a	2007-02-15 02:26:10 +0000	[diff] [blame]	3261	<dt><tt>VectorType</tt></dt>
Reid Spencer	485bad1	2007-02-15 03:07:05 +0000	[diff] [blame]	3262	<dd>Subclass of SequentialType for vector types. A
				3263	vector type is similar to an ArrayType but is distinguished because it is
Benjamin Kramer	8040cd3	2009-10-12 14:46:08 +0000	[diff] [blame]	3264	a first class type whereas ArrayType is not. Vector types are used for
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3265	vector operations and are usually small vectors of of an integer or floating
				3266	point type.</dd>
				3267	<dt><tt>StructType</tt></dt>
				3268	<dd>Subclass of DerivedTypes for struct types.</dd>
Duncan Sands	8036ca4	2007-03-30 12:22:09 +0000	[diff] [blame]	3269	<dt><tt><a name="FunctionType">FunctionType</a></tt></dt>
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3270	<dd>Subclass of DerivedTypes for function types.
				3271	<ul>
Dan Gohman	4bb31bf	2010-03-30 20:04:57 +0000	[diff] [blame]	3272	<li><tt>bool isVarArg() const</tt>: Returns true if it's a vararg
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3273	function</li>
				3274	<li><tt> const Type * getReturnType() const</tt>: Returns the
				3275	return type of the function.</li>
				3276	<li><tt>const Type * getParamType (unsigned i)</tt>: Returns
				3277	the type of the ith parameter.</li>
				3278	<li><tt> const unsigned getNumParams() const</tt>: Returns the
				3279	number of formal parameters.</li>
				3280	</ul>
				3281	</dd>
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3282	</dl>
				3283	</div>
				3284
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3285	</div>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3286
				3287	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3288	<h3>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3289	<a name="Module">The <tt>Module</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3290	</h3>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3291
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3292	<div>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3293
				3294	<p><tt>#include "<a
				3295	href="/doxygen/Module_8h-source.html">llvm/Module.h</a>"</tt><br> doxygen info:
				3296	<a href="/doxygen/classllvm_1_1Module.html">Module Class</a></p>
				3297
				3298	<p>The <tt>Module</tt> class represents the top level structure present in LLVM
				3299	programs. An LLVM module is effectively either a translation unit of the
				3300	original program or a combination of several translation units merged by the
				3301	linker. The <tt>Module</tt> class keeps track of a list of <a
				3302	href="#Function"><tt>Function</tt></a>s, a list of <a
				3303	href="#GlobalVariable"><tt>GlobalVariable</tt></a>s, and a <a
				3304	href="#SymbolTable"><tt>SymbolTable</tt></a>. Additionally, it contains a few
				3305	helpful member functions that try to make common operations easy.</p>
				3306
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3307	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3308	<h4>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3309	<a name="m_Module">Important Public Members of the <tt>Module</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3310	</h4>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3311
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3312	<div>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3313
				3314	<ul>
NAKAMURA Takumi	9c55f59	2012-03-27 11:25:16 +0000	[diff] [blame]	3315	<li><tt>Module::Module(std::string name = "")</tt>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3316
NAKAMURA Takumi	9c55f59	2012-03-27 11:25:16 +0000	[diff] [blame]	3317	<p>Constructing a <a href="#Module">Module</a> is easy. You can optionally
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3318	provide a name for it (probably based on the name of the translation unit).</p>
NAKAMURA Takumi	9c55f59	2012-03-27 11:25:16 +0000	[diff] [blame]	3319	</li>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3320
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3321	<li><tt>Module::iterator</tt> - Typedef for function list iterator<br>
				3322	<tt>Module::const_iterator</tt> - Typedef for const_iterator.<br>
				3323
				3324	<tt>begin()</tt>, <tt>end()</tt>
				3325	<tt>size()</tt>, <tt>empty()</tt>
				3326
				3327	<p>These are forwarding methods that make it easy to access the contents of
				3328	a <tt>Module</tt> object's <a href="#Function"><tt>Function</tt></a>
				3329	list.</p></li>
				3330
				3331	<li><tt>Module::FunctionListType &getFunctionList()</tt>
				3332
				3333	<p> Returns the list of <a href="#Function"><tt>Function</tt></a>s. This is
				3334	necessary to use when you need to update the list or perform a complex
				3335	action that doesn't have a forwarding method.</p>
				3336
				3337	<p><!-- Global Variable --></p></li>
				3338	</ul>
				3339
				3340	<hr>
				3341
				3342	<ul>
				3343	<li><tt>Module::global_iterator</tt> - Typedef for global variable list iterator<br>
				3344
				3345	<tt>Module::const_global_iterator</tt> - Typedef for const_iterator.<br>
				3346
				3347	<tt>global_begin()</tt>, <tt>global_end()</tt>
				3348	<tt>global_size()</tt>, <tt>global_empty()</tt>
				3349
				3350	<p> These are forwarding methods that make it easy to access the contents of
				3351	a <tt>Module</tt> object's <a
				3352	href="#GlobalVariable"><tt>GlobalVariable</tt></a> list.</p></li>
				3353
				3354	<li><tt>Module::GlobalListType &getGlobalList()</tt>
				3355
				3356	<p>Returns the list of <a
				3357	href="#GlobalVariable"><tt>GlobalVariable</tt></a>s. This is necessary to
				3358	use when you need to update the list or perform a complex action that
				3359	doesn't have a forwarding method.</p>
				3360
				3361	<p><!-- Symbol table stuff --> </p></li>
				3362	</ul>
				3363
				3364	<hr>
				3365
				3366	<ul>
				3367	<li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
				3368
				3369	<p>Return a reference to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
				3370	for this <tt>Module</tt>.</p>
				3371
				3372	<p><!-- Convenience methods --></p></li>
				3373	</ul>
				3374
				3375	<hr>
				3376
				3377	<ul>
Chad Rosier	c5171d5	2012-07-06 17:15:03 +0000	[diff] [blame]	3378
				3379	<li><tt><a href="#Function">Function</a> *getFunction(StringRef Name) const
				3380	</tt>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3381
				3382	<p>Look up the specified function in the <tt>Module</tt> <a
				3383	href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, return
				3384	<tt>null</tt>.</p></li>
				3385
				3386	<li><tt><a href="#Function">Function</a> *getOrInsertFunction(const
				3387	std::string &Name, const <a href="#FunctionType">FunctionType</a> *T)</tt>
				3388
				3389	<p>Look up the specified function in the <tt>Module</tt> <a
				3390	href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, add an
				3391	external declaration for the function and return it.</p></li>
				3392
				3393	<li><tt>std::string getTypeName(const <a href="#Type">Type</a> *Ty)</tt>
				3394
				3395	<p>If there is at least one entry in the <a
				3396	href="#SymbolTable"><tt>SymbolTable</tt></a> for the specified <a
				3397	href="#Type"><tt>Type</tt></a>, return it. Otherwise return the empty
				3398	string.</p></li>
				3399
				3400	<li><tt>bool addTypeName(const std::string &Name, const <a
				3401	href="#Type">Type</a> *Ty)</tt>
				3402
				3403	<p>Insert an entry in the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
				3404	mapping <tt>Name</tt> to <tt>Ty</tt>. If there is already an entry for this
				3405	name, true is returned and the <a
				3406	href="#SymbolTable"><tt>SymbolTable</tt></a> is not modified.</p></li>
				3407	</ul>
				3408
				3409	</div>
				3410
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3411	</div>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3412
Reid Spencer	303c4b4	2007-01-12 17:26:25 +0000	[diff] [blame]	3413	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3414	<h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3415	<a name="Value">The <tt>Value</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3416	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3417
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3418	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3419
				3420	<p><tt>#include "<a href="/doxygen/Value_8h-source.html">llvm/Value.h</a>"</tt>
				3421	<br>
Chris Lattner	0081517	2007-01-04 22:01:45 +0000	[diff] [blame]	3422	doxygen info: <a href="/doxygen/classllvm_1_1Value.html">Value Class</a></p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3423
				3424	<p>The <tt>Value</tt> class is the most important class in the LLVM Source
				3425	base. It represents a typed value that may be used (among other things) as an
				3426	operand to an instruction. There are many different types of <tt>Value</tt>s,
				3427	such as <a href="#Constant"><tt>Constant</tt></a>s,<a
				3428	href="#Argument"><tt>Argument</tt></a>s. Even <a
				3429	href="#Instruction"><tt>Instruction</tt></a>s and <a
				3430	href="#Function"><tt>Function</tt></a>s are <tt>Value</tt>s.</p>
				3431
				3432	<p>A particular <tt>Value</tt> may be used many times in the LLVM representation
				3433	for a program. For example, an incoming argument to a function (represented
				3434	with an instance of the <a href="#Argument">Argument</a> class) is "used" by
				3435	every instruction in the function that references the argument. To keep track
				3436	of this relationship, the <tt>Value</tt> class keeps a list of all of the <a
				3437	href="#User"><tt>User</tt></a>s that is using it (the <a
				3438	href="#User"><tt>User</tt></a> class is a base class for all nodes in the LLVM
				3439	graph that can refer to <tt>Value</tt>s). This use list is how LLVM represents
				3440	def-use information in the program, and is accessible through the <tt>use_</tt>*
				3441	methods, shown below.</p>
				3442
				3443	<p>Because LLVM is a typed representation, every LLVM <tt>Value</tt> is typed,
				3444	and this <a href="#Type">Type</a> is available through the <tt>getType()</tt>
				3445	method. In addition, all LLVM values can be named. The "name" of the
				3446	<tt>Value</tt> is a symbolic string printed in the LLVM code:</p>
				3447
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	3448	<div class="doc_code">
				3449	<pre>
Reid Spencer	06565dc	2007-01-12 17:11:23 +0000	[diff] [blame]	3450	%<b>foo</b> = add i32 1, 2
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	3451	</pre>
				3452	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3453
Duncan Sands	8036ca4	2007-03-30 12:22:09 +0000	[diff] [blame]	3454	<p><a name="nameWarning">The name of this instruction is "foo".</a> <b>NOTE</b>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3455	that the name of any value may be missing (an empty string), so names should
				3456	<b>ONLY</b> be used for debugging (making the source code easier to read,
				3457	debugging printouts), they should not be used to keep track of values or map
				3458	between them. For this purpose, use a <tt>std::map</tt> of pointers to the
				3459	<tt>Value</tt> itself instead.</p>
				3460
				3461	<p>One important aspect of LLVM is that there is no distinction between an SSA
				3462	variable and the operation that produces it. Because of this, any reference to
				3463	the value produced by an instruction (or the value available as an incoming
Chris Lattner	d5fc4fc	2004-03-18 14:58:55 +0000	[diff] [blame]	3464	argument, for example) is represented as a direct pointer to the instance of
				3465	the class that
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3466	represents this value. Although this may take some getting used to, it
				3467	simplifies the representation and makes it easier to manipulate.</p>
				3468
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3469	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3470	<h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3471	<a name="m_Value">Important Public Members of the <tt>Value</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3472	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3473
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3474	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3475
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3476	<ul>
				3477	<li><tt>Value::use_iterator</tt> - Typedef for iterator over the
				3478	use-list<br>
Gabor Greif	bbbf9a2	2010-03-26 19:59:25 +0000	[diff] [blame]	3479	<tt>Value::const_use_iterator</tt> - Typedef for const_iterator over
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3480	the use-list<br>
				3481	<tt>unsigned use_size()</tt> - Returns the number of users of the
				3482	value.<br>
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	3483	<tt>bool use_empty()</tt> - Returns true if there are no users.<br>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3484	<tt>use_iterator use_begin()</tt> - Get an iterator to the start of
				3485	the use-list.<br>
				3486	<tt>use_iterator use_end()</tt> - Get an iterator to the end of the
				3487	use-list.<br>
				3488	<tt><a href="#User">User</a> *use_back()</tt> - Returns the last
				3489	element in the list.
				3490	<p> These methods are the interface to access the def-use
				3491	information in LLVM. As with all other iterators in LLVM, the naming
				3492	conventions follow the conventions defined by the <a href="#stl">STL</a>.</p>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3493	</li>
				3494	<li><tt><a href="#Type">Type</a> *getType() const</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3495	<p>This method returns the Type of the Value.</p>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3496	</li>
				3497	<li><tt>bool hasName() const</tt><br>
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	3498	<tt>std::string getName() const</tt><br>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3499	<tt>void setName(const std::string &Name)</tt>
				3500	<p> This family of methods is used to access and assign a name to a <tt>Value</tt>,
				3501	be aware of the <a href="#nameWarning">precaution above</a>.</p>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3502	</li>
				3503	<li><tt>void replaceAllUsesWith(Value *V)</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3504
				3505	<p>This method traverses the use list of a <tt>Value</tt> changing all <a
				3506	href="#User"><tt>User</tt>s</a> of the current value to refer to
				3507	"<tt>V</tt>" instead. For example, if you detect that an instruction always
				3508	produces a constant value (for example through constant folding), you can
				3509	replace all uses of the instruction with the constant like this:</p>
				3510
Bill Wendling	3cd5ca6	2006-10-11 06:30:10 +0000	[diff] [blame]	3511	<div class="doc_code">
				3512	<pre>
				3513	Inst->replaceAllUsesWith(ConstVal);
				3514	</pre>
				3515	</div>
				3516
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3517	</ul>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3518
				3519	</div>
				3520
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3521	</div>
				3522
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3523	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3524	<h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3525	<a name="User">The <tt>User</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3526	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3527
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3528	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3529
				3530	<p>
				3531	<tt>#include "<a href="/doxygen/User_8h-source.html">llvm/User.h</a>"</tt><br>
Misha Brukman	384047f	2004-06-03 23:29:12 +0000	[diff] [blame]	3532	doxygen info: <a href="/doxygen/classllvm_1_1User.html">User Class</a><br>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3533	Superclass: <a href="#Value"><tt>Value</tt></a></p>
				3534
				3535	<p>The <tt>User</tt> class is the common base class of all LLVM nodes that may
				3536	refer to <a href="#Value"><tt>Value</tt></a>s. It exposes a list of "Operands"
				3537	that are all of the <a href="#Value"><tt>Value</tt></a>s that the User is
				3538	referring to. The <tt>User</tt> class itself is a subclass of
				3539	<tt>Value</tt>.</p>
				3540
				3541	<p>The operands of a <tt>User</tt> point directly to the LLVM <a
				3542	href="#Value"><tt>Value</tt></a> that it refers to. Because LLVM uses Static
				3543	Single Assignment (SSA) form, there can only be one definition referred to,
				3544	allowing this direct connection. This connection provides the use-def
				3545	information in LLVM.</p>
				3546
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3547	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3548	<h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3549	<a name="m_User">Important Public Members of the <tt>User</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3550	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3551
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3552	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3553
				3554	<p>The <tt>User</tt> class exposes the operand list in two ways: through
				3555	an index access interface and through an iterator based interface.</p>
				3556
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3557	<ul>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3558	<li><tt>Value *getOperand(unsigned i)</tt><br>
				3559	<tt>unsigned getNumOperands()</tt>
				3560	<p> These two methods expose the operands of the <tt>User</tt> in a
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3561	convenient form for direct access.</p></li>
				3562
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3563	<li><tt>User::op_iterator</tt> - Typedef for iterator over the operand
				3564	list<br>
Chris Lattner	5836082	2005-01-17 00:12:04 +0000	[diff] [blame]	3565	<tt>op_iterator op_begin()</tt> - Get an iterator to the start of
				3566	the operand list.<br>
				3567	<tt>op_iterator op_end()</tt> - Get an iterator to the end of the
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3568	operand list.
				3569	<p> Together, these methods make up the iterator based interface to
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3570	the operands of a <tt>User</tt>.</p></li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3571	</ul>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3572
				3573	</div>
				3574
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3575	</div>
				3576
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3577	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3578	<h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3579	<a name="Instruction">The <tt>Instruction</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3580	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3581
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3582	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3583
				3584	<p><tt>#include "</tt><tt><a
				3585	href="/doxygen/Instruction_8h-source.html">llvm/Instruction.h</a>"</tt><br>
Misha Brukman	31ca1de	2004-06-03 23:35:54 +0000	[diff] [blame]	3586	doxygen info: <a href="/doxygen/classllvm_1_1Instruction.html">Instruction Class</a><br>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3587	Superclasses: <a href="#User"><tt>User</tt></a>, <a
				3588	href="#Value"><tt>Value</tt></a></p>
				3589
				3590	<p>The <tt>Instruction</tt> class is the common base class for all LLVM
				3591	instructions. It provides only a few methods, but is a very commonly used
				3592	class. The primary data tracked by the <tt>Instruction</tt> class itself is the
				3593	opcode (instruction type) and the parent <a
				3594	href="#BasicBlock"><tt>BasicBlock</tt></a> the <tt>Instruction</tt> is embedded
				3595	into. To represent a specific type of instruction, one of many subclasses of
				3596	<tt>Instruction</tt> are used.</p>
				3597
				3598	<p> Because the <tt>Instruction</tt> class subclasses the <a
				3599	href="#User"><tt>User</tt></a> class, its operands can be accessed in the same
				3600	way as for other <a href="#User"><tt>User</tt></a>s (with the
				3601	<tt>getOperand()</tt>/<tt>getNumOperands()</tt> and
				3602	<tt>op_begin()</tt>/<tt>op_end()</tt> methods).</p> <p> An important file for
				3603	the <tt>Instruction</tt> class is the <tt>llvm/Instruction.def</tt> file. This
				3604	file contains some meta-data about the various different types of instructions
				3605	in LLVM. It describes the enum values that are used as opcodes (for example
Reid Spencer	c92d25d	2006-12-19 19:47:19 +0000	[diff] [blame]	3606	<tt>Instruction::Add</tt> and <tt>Instruction::ICmp</tt>), as well as the
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3607	concrete sub-classes of <tt>Instruction</tt> that implement the instruction (for
				3608	example <tt><a href="#BinaryOperator">BinaryOperator</a></tt> and <tt><a
Reid Spencer	c92d25d	2006-12-19 19:47:19 +0000	[diff] [blame]	3609	href="#CmpInst">CmpInst</a></tt>). Unfortunately, the use of macros in
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3610	this file confuses doxygen, so these enum values don't show up correctly in the
Misha Brukman	31ca1de	2004-06-03 23:35:54 +0000	[diff] [blame]	3611	<a href="/doxygen/classllvm_1_1Instruction.html">doxygen output</a>.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3612
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3613	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3614	<h4>
				3615	<a name="s_Instruction">
				3616	Important Subclasses of the <tt>Instruction</tt> class
				3617	</a>
				3618	</h4>
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3619	<div>
Reid Spencer	c92d25d	2006-12-19 19:47:19 +0000	[diff] [blame]	3620	<ul>
				3621	<li><tt><a name="BinaryOperator">BinaryOperator</a></tt>
				3622	<p>This subclasses represents all two operand instructions whose operands
				3623	must be the same type, except for the comparison instructions.</p></li>
				3624	<li><tt><a name="CastInst">CastInst</a></tt>
				3625	<p>This subclass is the parent of the 12 casting instructions. It provides
				3626	common operations on cast instructions.</p>
				3627	<li><tt><a name="CmpInst">CmpInst</a></tt>
				3628	<p>This subclass respresents the two comparison instructions,
				3629	<a href="LangRef.html#i_icmp">ICmpInst</a> (integer opreands), and
				3630	<a href="LangRef.html#i_fcmp">FCmpInst</a> (floating point operands).</p>
				3631	<li><tt><a name="TerminatorInst">TerminatorInst</a></tt>
				3632	<p>This subclass is the parent of all terminator instructions (those which
				3633	can terminate a block).</p>
				3634	</ul>
				3635	</div>
				3636
				3637	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3638	<h4>
				3639	<a name="m_Instruction">
				3640	Important Public Members of the <tt>Instruction</tt> class
				3641	</a>
				3642	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3643
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3644	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3645
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3646	<ul>
				3647	<li><tt><a href="#BasicBlock">BasicBlock</a> *getParent()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3648	<p>Returns the <a href="#BasicBlock"><tt>BasicBlock</tt></a> that
				3649	this <tt>Instruction</tt> is embedded into.</p></li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3650	<li><tt>bool mayWriteToMemory()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3651	<p>Returns true if the instruction writes to memory, i.e. it is a
				3652	<tt>call</tt>,<tt>free</tt>,<tt>invoke</tt>, or <tt>store</tt>.</p></li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3653	<li><tt>unsigned getOpcode()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3654	<p>Returns the opcode for the <tt>Instruction</tt>.</p></li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3655	<li><tt><a href="#Instruction">Instruction</a> *clone() const</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3656	<p>Returns another instance of the specified instruction, identical
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3657	in all ways to the original except that the instruction has no parent
				3658	(ie it's not embedded into a <a href="#BasicBlock"><tt>BasicBlock</tt></a>),
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3659	and it has no name</p></li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3660	</ul>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3661
				3662	</div>
				3663
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3664	</div>
				3665
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3666	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3667	<h3>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3668	<a name="Constant">The <tt>Constant</tt> class and subclasses</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3669	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3670
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3671	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3672
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3673	<p>Constant represents a base class for different types of constants. It
				3674	is subclassed by ConstantInt, ConstantArray, etc. for representing
				3675	the various types of Constants. <a href="#GlobalValue">GlobalValue</a> is also
				3676	a subclass, which represents the address of a global variable or function.
				3677	</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3678
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3679	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3680	<h4>Important Subclasses of Constant</h4>
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3681	<div>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3682	<ul>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3683	<li>ConstantInt : This subclass of Constant represents an integer constant of
				3684	any width.
				3685	<ul>
Reid Spencer	97b4ee3	2007-03-01 21:05:33 +0000	[diff] [blame]	3686	<li><tt>const APInt& getValue() const</tt>: Returns the underlying
				3687	value of this constant, an APInt value.</li>
				3688	<li><tt>int64_t getSExtValue() const</tt>: Converts the underlying APInt
				3689	value to an int64_t via sign extension. If the value (not the bit width)
				3690	of the APInt is too large to fit in an int64_t, an assertion will result.
				3691	For this reason, use of this method is discouraged.</li>
				3692	<li><tt>uint64_t getZExtValue() const</tt>: Converts the underlying APInt
				3693	value to a uint64_t via zero extension. IF the value (not the bit width)
				3694	of the APInt is too large to fit in a uint64_t, an assertion will result.
Reid Spencer	4474d87	2007-03-02 01:31:31 +0000	[diff] [blame]	3695	For this reason, use of this method is discouraged.</li>
Reid Spencer	97b4ee3	2007-03-01 21:05:33 +0000	[diff] [blame]	3696	<li><tt>static ConstantInt* get(const APInt& Val)</tt>: Returns the
				3697	ConstantInt object that represents the value provided by <tt>Val</tt>.
				3698	The type is implied as the IntegerType that corresponds to the bit width
				3699	of <tt>Val</tt>.</li>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3700	<li><tt>static ConstantInt* get(const Type *Ty, uint64_t Val)</tt>:
				3701	Returns the ConstantInt object that represents the value provided by
				3702	<tt>Val</tt> for integer type <tt>Ty</tt>.</li>
				3703	</ul>
				3704	</li>
				3705	<li>ConstantFP : This class represents a floating point constant.
				3706	<ul>
				3707	<li><tt>double getValue() const</tt>: Returns the underlying value of
				3708	this constant. </li>
				3709	</ul>
				3710	</li>
				3711	<li>ConstantArray : This represents a constant array.
				3712	<ul>
				3713	<li><tt>const std::vector<Use> &getValues() const</tt>: Returns
				3714	a vector of component constants that makeup this array. </li>
				3715	</ul>
				3716	</li>
				3717	<li>ConstantStruct : This represents a constant struct.
				3718	<ul>
				3719	<li><tt>const std::vector<Use> &getValues() const</tt>: Returns
				3720	a vector of component constants that makeup this array. </li>
				3721	</ul>
				3722	</li>
				3723	<li>GlobalValue : This represents either a global variable or a function. In
				3724	either case, the value is a constant fixed address (after linking).
				3725	</li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3726	</ul>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3727	</div>
				3728
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3729	</div>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	3730
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3731	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3732	<h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3733	<a name="GlobalValue">The <tt>GlobalValue</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3734	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3735
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3736	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3737
				3738	<p><tt>#include "<a
				3739	href="/doxygen/GlobalValue_8h-source.html">llvm/GlobalValue.h</a>"</tt><br>
Misha Brukman	384047f	2004-06-03 23:29:12 +0000	[diff] [blame]	3740	doxygen info: <a href="/doxygen/classllvm_1_1GlobalValue.html">GlobalValue
				3741	Class</a><br>
Reid Spencer	be5e85e	2006-04-14 14:11:48 +0000	[diff] [blame]	3742	Superclasses: <a href="#Constant"><tt>Constant</tt></a>,
				3743	<a href="#User"><tt>User</tt></a>, <a href="#Value"><tt>Value</tt></a></p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3744
				3745	<p>Global values (<a href="#GlobalVariable"><tt>GlobalVariable</tt></a>s or <a
				3746	href="#Function"><tt>Function</tt></a>s) are the only LLVM values that are
				3747	visible in the bodies of all <a href="#Function"><tt>Function</tt></a>s.
				3748	Because they are visible at global scope, they are also subject to linking with
				3749	other globals defined in different translation units. To control the linking
				3750	process, <tt>GlobalValue</tt>s know their linkage rules. Specifically,
				3751	<tt>GlobalValue</tt>s know whether they have internal or external linkage, as
Reid Spencer	8b2da7a	2004-07-18 13:10:31 +0000	[diff] [blame]	3752	defined by the <tt>LinkageTypes</tt> enumeration.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3753
				3754	<p>If a <tt>GlobalValue</tt> has internal linkage (equivalent to being
				3755	<tt>static</tt> in C), it is not visible to code outside the current translation
				3756	unit, and does not participate in linking. If it has external linkage, it is
				3757	visible to external code, and does participate in linking. In addition to
				3758	linkage information, <tt>GlobalValue</tt>s keep track of which <a
				3759	href="#Module"><tt>Module</tt></a> they are currently part of.</p>
				3760
				3761	<p>Because <tt>GlobalValue</tt>s are memory objects, they are always referred to
				3762	by their <b>address</b>. As such, the <a href="#Type"><tt>Type</tt></a> of a
				3763	global is always a pointer to its contents. It is important to remember this
				3764	when using the <tt>GetElementPtrInst</tt> instruction because this pointer must
				3765	be dereferenced first. For example, if you have a <tt>GlobalVariable</tt> (a
				3766	subclass of <tt>GlobalValue)</tt> that is an array of 24 ints, type <tt>[24 x
Reid Spencer	06565dc	2007-01-12 17:11:23 +0000	[diff] [blame]	3767	i32]</tt>, then the <tt>GlobalVariable</tt> is a pointer to that array. Although
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3768	the address of the first element of this array and the value of the
				3769	<tt>GlobalVariable</tt> are the same, they have different types. The
Reid Spencer	06565dc	2007-01-12 17:11:23 +0000	[diff] [blame]	3770	<tt>GlobalVariable</tt>'s type is <tt>[24 x i32]</tt>. The first element's type
				3771	is <tt>i32.</tt> Because of this, accessing a global value requires you to
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3772	dereference the pointer with <tt>GetElementPtrInst</tt> first, then its elements
				3773	can be accessed. This is explained in the <a href="LangRef.html#globalvars">LLVM
				3774	Language Reference Manual</a>.</p>
				3775
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3776	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3777	<h4>
				3778	<a name="m_GlobalValue">
				3779	Important Public Members of the <tt>GlobalValue</tt> class
				3780	</a>
				3781	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3782
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3783	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3784
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3785	<ul>
				3786	<li><tt>bool hasInternalLinkage() const</tt><br>
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	3787	<tt>bool hasExternalLinkage() const</tt><br>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3788	<tt>void setInternalLinkage(bool HasInternalLinkage)</tt>
				3789	<p> These methods manipulate the linkage characteristics of the <tt>GlobalValue</tt>.</p>
				3790	<p> </p>
				3791	</li>
				3792	<li><tt><a href="#Module">Module</a> *getParent()</tt>
				3793	<p> This returns the <a href="#Module"><tt>Module</tt></a> that the
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3794	GlobalValue is currently embedded into.</p></li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3795	</ul>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3796
				3797	</div>
				3798
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3799	</div>
				3800
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3801	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3802	<h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3803	<a name="Function">The <tt>Function</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3804	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3805
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3806	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3807
				3808	<p><tt>#include "<a
				3809	href="/doxygen/Function_8h-source.html">llvm/Function.h</a>"</tt><br> doxygen
Misha Brukman	31ca1de	2004-06-03 23:35:54 +0000	[diff] [blame]	3810	info: <a href="/doxygen/classllvm_1_1Function.html">Function Class</a><br>
Reid Spencer	be5e85e	2006-04-14 14:11:48 +0000	[diff] [blame]	3811	Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
				3812	<a href="#Constant"><tt>Constant</tt></a>,
				3813	<a href="#User"><tt>User</tt></a>,
				3814	<a href="#Value"><tt>Value</tt></a></p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3815
				3816	<p>The <tt>Function</tt> class represents a single procedure in LLVM. It is
Torok Edwin	8746929	2009-10-12 13:37:29 +0000	[diff] [blame]	3817	actually one of the more complex classes in the LLVM hierarchy because it must
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3818	keep track of a large amount of data. The <tt>Function</tt> class keeps track
Reid Spencer	be5e85e	2006-04-14 14:11:48 +0000	[diff] [blame]	3819	of a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, a list of formal
				3820	<a href="#Argument"><tt>Argument</tt></a>s, and a
				3821	<a href="#SymbolTable"><tt>SymbolTable</tt></a>.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3822
				3823	<p>The list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s is the most
				3824	commonly used part of <tt>Function</tt> objects. The list imposes an implicit
				3825	ordering of the blocks in the function, which indicate how the code will be
Benjamin Kramer	8040cd3	2009-10-12 14:46:08 +0000	[diff] [blame]	3826	laid out by the backend. Additionally, the first <a
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3827	href="#BasicBlock"><tt>BasicBlock</tt></a> is the implicit entry node for the
				3828	<tt>Function</tt>. It is not legal in LLVM to explicitly branch to this initial
				3829	block. There are no implicit exit nodes, and in fact there may be multiple exit
				3830	nodes from a single <tt>Function</tt>. If the <a
				3831	href="#BasicBlock"><tt>BasicBlock</tt></a> list is empty, this indicates that
				3832	the <tt>Function</tt> is actually a function declaration: the actual body of the
				3833	function hasn't been linked in yet.</p>
				3834
				3835	<p>In addition to a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, the
				3836	<tt>Function</tt> class also keeps track of the list of formal <a
				3837	href="#Argument"><tt>Argument</tt></a>s that the function receives. This
				3838	container manages the lifetime of the <a href="#Argument"><tt>Argument</tt></a>
				3839	nodes, just like the <a href="#BasicBlock"><tt>BasicBlock</tt></a> list does for
				3840	the <a href="#BasicBlock"><tt>BasicBlock</tt></a>s.</p>
				3841
				3842	<p>The <a href="#SymbolTable"><tt>SymbolTable</tt></a> is a very rarely used
				3843	LLVM feature that is only used when you have to look up a value by name. Aside
				3844	from that, the <a href="#SymbolTable"><tt>SymbolTable</tt></a> is used
				3845	internally to make sure that there are not conflicts between the names of <a
				3846	href="#Instruction"><tt>Instruction</tt></a>s, <a
				3847	href="#BasicBlock"><tt>BasicBlock</tt></a>s, or <a
				3848	href="#Argument"><tt>Argument</tt></a>s in the function body.</p>
				3849
Reid Spencer	8b2da7a	2004-07-18 13:10:31 +0000	[diff] [blame]	3850	<p>Note that <tt>Function</tt> is a <a href="#GlobalValue">GlobalValue</a>
				3851	and therefore also a <a href="#Constant">Constant</a>. The value of the function
				3852	is its address (after linking) which is guaranteed to be constant.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3853
				3854	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3855	<h4>
				3856	<a name="m_Function">
				3857	Important Public Members of the <tt>Function</tt> class
				3858	</a>
				3859	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3860
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3861	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3862
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3863	<ul>
				3864	<li><tt>Function(const </tt><tt><a href="#FunctionType">FunctionType</a>
Chris Lattner	ac479e5	2004-08-04 05:10:48 +0000	[diff] [blame]	3865	Ty, LinkageTypes Linkage, const std::string &N = "", Module Parent = 0)</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3866
				3867	<p>Constructor used when you need to create new <tt>Function</tt>s to add
Sylvestre Ledru	c8e41c5	2012-07-23 08:51:15 +0000	[diff] [blame]	3868	the program. The constructor must specify the type of the function to
Chris Lattner	ac479e5	2004-08-04 05:10:48 +0000	[diff] [blame]	3869	create and what type of linkage the function should have. The <a
				3870	href="#FunctionType"><tt>FunctionType</tt></a> argument
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3871	specifies the formal arguments and return value for the function. The same
Duncan Sands	8036ca4	2007-03-30 12:22:09 +0000	[diff] [blame]	3872	<a href="#FunctionType"><tt>FunctionType</tt></a> value can be used to
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3873	create multiple functions. The <tt>Parent</tt> argument specifies the Module
				3874	in which the function is defined. If this argument is provided, the function
				3875	will automatically be inserted into that module's list of
				3876	functions.</p></li>
				3877
Chris Lattner	62810e3	2008-11-25 18:34:50 +0000	[diff] [blame]	3878	<li><tt>bool isDeclaration()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3879
				3880	<p>Return whether or not the <tt>Function</tt> has a body defined. If the
				3881	function is "external", it does not have a body, and thus must be resolved
				3882	by linking with a function defined in a different translation unit.</p></li>
				3883
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3884	<li><tt>Function::iterator</tt> - Typedef for basic block list iterator<br>
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	3885	<tt>Function::const_iterator</tt> - Typedef for const_iterator.<br>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3886
Chris Lattner	77d6924	2005-03-15 05:19:20 +0000	[diff] [blame]	3887	<tt>begin()</tt>, <tt>end()</tt>
				3888	<tt>size()</tt>, <tt>empty()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3889
				3890	<p>These are forwarding methods that make it easy to access the contents of
				3891	a <tt>Function</tt> object's <a href="#BasicBlock"><tt>BasicBlock</tt></a>
				3892	list.</p></li>
				3893
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3894	<li><tt>Function::BasicBlockListType &getBasicBlockList()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3895
				3896	<p>Returns the list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s. This
				3897	is necessary to use when you need to update the list or perform a complex
				3898	action that doesn't have a forwarding method.</p></li>
				3899
Chris Lattner	89cc265	2005-03-15 04:48:32 +0000	[diff] [blame]	3900	<li><tt>Function::arg_iterator</tt> - Typedef for the argument list
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3901	iterator<br>
Chris Lattner	89cc265	2005-03-15 04:48:32 +0000	[diff] [blame]	3902	<tt>Function::const_arg_iterator</tt> - Typedef for const_iterator.<br>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3903
Chris Lattner	77d6924	2005-03-15 05:19:20 +0000	[diff] [blame]	3904	<tt>arg_begin()</tt>, <tt>arg_end()</tt>
Chris Lattner	89cc265	2005-03-15 04:48:32 +0000	[diff] [blame]	3905	<tt>arg_size()</tt>, <tt>arg_empty()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3906
				3907	<p>These are forwarding methods that make it easy to access the contents of
				3908	a <tt>Function</tt> object's <a href="#Argument"><tt>Argument</tt></a>
				3909	list.</p></li>
				3910
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3911	<li><tt>Function::ArgumentListType &getArgumentList()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3912
				3913	<p>Returns the list of <a href="#Argument"><tt>Argument</tt></a>s. This is
				3914	necessary to use when you need to update the list or perform a complex
				3915	action that doesn't have a forwarding method.</p></li>
				3916
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3917	<li><tt><a href="#BasicBlock">BasicBlock</a> &getEntryBlock()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3918
				3919	<p>Returns the entry <a href="#BasicBlock"><tt>BasicBlock</tt></a> for the
				3920	function. Because the entry block for the function is always the first
				3921	block, this returns the first block of the <tt>Function</tt>.</p></li>
				3922
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3923	<li><tt><a href="#Type">Type</a> *getReturnType()</tt><br>
				3924	<tt><a href="#FunctionType">FunctionType</a> *getFunctionType()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3925
				3926	<p>This traverses the <a href="#Type"><tt>Type</tt></a> of the
				3927	<tt>Function</tt> and returns the return type of the function, or the <a
				3928	href="#FunctionType"><tt>FunctionType</tt></a> of the actual
				3929	function.</p></li>
				3930
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3931	<li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3932
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3933	<p> Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3934	for this <tt>Function</tt>.</p></li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3935	</ul>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3936
				3937	</div>
				3938
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3939	</div>
				3940
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3941	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3942	<h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3943	<a name="GlobalVariable">The <tt>GlobalVariable</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3944	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3945
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3946	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3947
				3948	<p><tt>#include "<a
				3949	href="/doxygen/GlobalVariable_8h-source.html">llvm/GlobalVariable.h</a>"</tt>
				3950	<br>
Tanya Lattner	a3da777	2004-06-22 08:02:25 +0000	[diff] [blame]	3951	doxygen info: <a href="/doxygen/classllvm_1_1GlobalVariable.html">GlobalVariable
Reid Spencer	be5e85e	2006-04-14 14:11:48 +0000	[diff] [blame]	3952	Class</a><br>
				3953	Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
				3954	<a href="#Constant"><tt>Constant</tt></a>,
				3955	<a href="#User"><tt>User</tt></a>,
				3956	<a href="#Value"><tt>Value</tt></a></p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3957
Benjamin Kramer	8040cd3	2009-10-12 14:46:08 +0000	[diff] [blame]	3958	<p>Global variables are represented with the (surprise surprise)
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3959	<tt>GlobalVariable</tt> class. Like functions, <tt>GlobalVariable</tt>s are also
				3960	subclasses of <a href="#GlobalValue"><tt>GlobalValue</tt></a>, and as such are
				3961	always referenced by their address (global values must live in memory, so their
Reid Spencer	be5e85e	2006-04-14 14:11:48 +0000	[diff] [blame]	3962	"name" refers to their constant address). See
				3963	<a href="#GlobalValue"><tt>GlobalValue</tt></a> for more on this. Global
				3964	variables may have an initial value (which must be a
				3965	<a href="#Constant"><tt>Constant</tt></a>), and if they have an initializer,
				3966	they may be marked as "constant" themselves (indicating that their contents
				3967	never change at runtime).</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3968
				3969	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	3970	<h4>
				3971	<a name="m_GlobalVariable">
				3972	Important Public Members of the <tt>GlobalVariable</tt> class
				3973	</a>
				3974	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3975
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	3976	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3977
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3978	<ul>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3979	<li><tt>GlobalVariable(const </tt><tt><a href="#Type">Type</a> *Ty, bool
				3980	isConstant, LinkageTypes& Linkage, <a href="#Constant">Constant</a>
				3981	Initializer = 0, const std::string &Name = "", Module Parent = 0)</tt>
				3982
				3983	<p>Create a new global variable of the specified type. If
				3984	<tt>isConstant</tt> is true then the global variable will be marked as
				3985	unchanging for the program. The Linkage parameter specifies the type of
Duncan Sands	667d4b8	2009-03-07 15:45:40 +0000	[diff] [blame]	3986	linkage (internal, external, weak, linkonce, appending) for the variable.
				3987	If the linkage is InternalLinkage, WeakAnyLinkage, WeakODRLinkage,
				3988	LinkOnceAnyLinkage or LinkOnceODRLinkage,  then the resultant
				3989	global variable will have internal linkage. AppendingLinkage concatenates
				3990	together all instances (in different translation units) of the variable
				3991	into a single variable but is only applicable to arrays.  See
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3992	the <a href="LangRef.html#modulestructure">LLVM Language Reference</a> for
				3993	further details on linkage types. Optionally an initializer, a name, and the
				3994	module to put the variable into may be specified for the global variable as
				3995	well.</p></li>
				3996
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	3997	<li><tt>bool isConstant() const</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	3998
				3999	<p>Returns true if this is a global variable that is known not to
				4000	be modified at runtime.</p></li>
				4001
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	4002	<li><tt>bool hasInitializer()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4003
				4004	<p>Returns true if this <tt>GlobalVariable</tt> has an intializer.</p></li>
				4005
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	4006	<li><tt><a href="#Constant">Constant</a> *getInitializer()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4007
Benjamin Kramer	8040cd3	2009-10-12 14:46:08 +0000	[diff] [blame]	4008	<p>Returns the initial value for a <tt>GlobalVariable</tt>. It is not legal
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4009	to call this method if there is no initializer.</p></li>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	4010	</ul>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4011
				4012	</div>
				4013
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	4014	</div>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4015
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4016	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	4017	<h3>
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4018	<a name="BasicBlock">The <tt>BasicBlock</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	4019	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4020
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	4021	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4022
				4023	<p><tt>#include "<a
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4024	href="/doxygen/BasicBlock_8h-source.html">llvm/BasicBlock.h</a>"</tt><br>
Stefanus Du Toit	24e0411	2009-06-17 21:12:26 +0000	[diff] [blame]	4025	doxygen info: <a href="/doxygen/classllvm_1_1BasicBlock.html">BasicBlock
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4026	Class</a><br>
				4027	Superclass: <a href="#Value"><tt>Value</tt></a></p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4028
Nick Lewycky	ccd279d	2011-02-17 02:19:22 +0000	[diff] [blame]	4029	<p>This class represents a single entry single exit section of the code,
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4030	commonly known as a basic block by the compiler community. The
				4031	<tt>BasicBlock</tt> class maintains a list of <a
				4032	href="#Instruction"><tt>Instruction</tt></a>s, which form the body of the block.
				4033	Matching the language definition, the last element of this list of instructions
				4034	is always a terminator instruction (a subclass of the <a
				4035	href="#TerminatorInst"><tt>TerminatorInst</tt></a> class).</p>
				4036
				4037	<p>In addition to tracking the list of instructions that make up the block, the
				4038	<tt>BasicBlock</tt> class also keeps track of the <a
				4039	href="#Function"><tt>Function</tt></a> that it is embedded into.</p>
				4040
				4041	<p>Note that <tt>BasicBlock</tt>s themselves are <a
				4042	href="#Value"><tt>Value</tt></a>s, because they are referenced by instructions
				4043	like branches and can go in the switch tables. <tt>BasicBlock</tt>s have type
				4044	<tt>label</tt>.</p>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4045
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4046	<!-- _______________________________________________________________________ -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	4047	<h4>
				4048	<a name="m_BasicBlock">
				4049	Important Public Members of the <tt>BasicBlock</tt> class
				4050	</a>
				4051	</h4>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4052
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	4053	<div>
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	4054	<ul>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4055
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4056	<li><tt>BasicBlock(const std::string &Name = "", </tt><tt><a
				4057	href="#Function">Function</a> *Parent = 0)</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4058
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4059	<p>The <tt>BasicBlock</tt> constructor is used to create new basic blocks for
				4060	insertion into a function. The constructor optionally takes a name for the new
				4061	block, and a <a href="#Function"><tt>Function</tt></a> to insert it into. If
				4062	the <tt>Parent</tt> parameter is specified, the new <tt>BasicBlock</tt> is
				4063	automatically inserted at the end of the specified <a
				4064	href="#Function"><tt>Function</tt></a>, if not specified, the BasicBlock must be
				4065	manually inserted into the <a href="#Function"><tt>Function</tt></a>.</p></li>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4066
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4067	<li><tt>BasicBlock::iterator</tt> - Typedef for instruction list iterator<br>
				4068	<tt>BasicBlock::const_iterator</tt> - Typedef for const_iterator.<br>
				4069	<tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
				4070	<tt>size()</tt>, <tt>empty()</tt>
				4071	STL-style functions for accessing the instruction list.
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4072
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4073	<p>These methods and typedefs are forwarding functions that have the same
				4074	semantics as the standard library methods of the same names. These methods
				4075	expose the underlying instruction list of a basic block in a way that is easy to
				4076	manipulate. To get the full complement of container operations (including
				4077	operations to update the list), you must use the <tt>getInstList()</tt>
				4078	method.</p></li>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4079
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4080	<li><tt>BasicBlock::InstListType &getInstList()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4081
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4082	<p>This method is used to get access to the underlying container that actually
				4083	holds the Instructions. This method must be used when there isn't a forwarding
				4084	function in the <tt>BasicBlock</tt> class for the operation that you would like
				4085	to perform. Because there are no forwarding functions for "updating"
				4086	operations, you need to use this if you want to update the contents of a
				4087	<tt>BasicBlock</tt>.</p></li>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4088
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4089	<li><tt><a href="#Function">Function</a> *getParent()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4090
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4091	<p> Returns a pointer to <a href="#Function"><tt>Function</tt></a> the block is
				4092	embedded into, or a null pointer if it is homeless.</p></li>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4093
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4094	<li><tt><a href="#TerminatorInst">TerminatorInst</a> *getTerminator()</tt>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4095
Chris Lattner	2b78d96	2007-02-03 20:02:25 +0000	[diff] [blame]	4096	<p> Returns a pointer to the terminator instruction that appears at the end of
				4097	the <tt>BasicBlock</tt>. If there is no terminator instruction, or if the last
				4098	instruction in the block is not a terminator, then a null pointer is
				4099	returned.</p></li>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4100
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4101	</ul>
				4102
				4103	</div>
				4104
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	4105	</div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4106
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4107	<!-- ======================================================================= -->
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	4108	<h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4109	<a name="Argument">The <tt>Argument</tt> class</a>
NAKAMURA Takumi	05d0265	2011-04-18 23:59:50 +0000	[diff] [blame]	4110	</h3>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4111
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	4112	<div>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4113
				4114	<p>This subclass of Value defines the interface for incoming formal
Chris Lattner	5836082	2005-01-17 00:12:04 +0000	[diff] [blame]	4115	arguments to a function. A Function maintains a list of its formal
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4116	arguments. An argument has a pointer to the parent Function.</p>
				4117
				4118	</div>
				4119
NAKAMURA Takumi	f5af6ad	2011-04-23 00:30:22 +0000	[diff] [blame]	4120	</div>
				4121
Chris Lattner	9355b47	2002-09-06 02:50:58 +0000	[diff] [blame]	4122	<!-- *********************************************************************** -->
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4123	<hr>
				4124	<address>
				4125	<a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman	4440870	2008-12-11 17:34:48 +0000	[diff] [blame]	4126	src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4127	<a href="http://validator.w3.org/check/referer"><img
Gabor Greif	a9c0f2b	2008-06-18 14:05:31 +0000	[diff] [blame]	4128	src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01 Strict"></a>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4129
				4130	<a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a> and
				4131	<a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
NAKAMURA Takumi	b9a3363	2011-04-09 02:13:37 +0000	[diff] [blame]	4132	<a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
Misha Brukman	13fd15c	2004-01-15 00:14:41 +0000	[diff] [blame]	4133	Last modified: $Date$
				4134	</address>
				4135
Chris Lattner	261efe9	2003-11-25 01:02:51 +0000	[diff] [blame]	4136	</body>
				4137	</html>