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

blob: 591694a983cf3513c76f540d8a68bf43de0d93d7 [file] [log] [blame]

Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1	<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
				2	"http://www.w3.org/TR/html4/strict.dtd">
				3	<html>
				4	<head>
				5	<title>LLVM Alias Analysis Infrastructure</title>
				6	<link rel="stylesheet" href="llvm.css" type="text/css">
				7	</head>
				8	<body>
				9
				10	<div class="doc_title">
				11	LLVM Alias Analysis Infrastructure
				12	</div>
				13
				14	<ol>
				15	<li><a href="#introduction">Introduction</a></li>
				16
				17	<li><a href="#overview"><tt>AliasAnalysis</tt> Class Overview</a>
				18	<ul>
				19	<li><a href="#pointers">Representation of Pointers</a></li>
				20	<li><a href="#alias">The <tt>alias</tt> method</a></li>
				21	<li><a href="#ModRefInfo">The <tt>getModRefInfo</tt> methods</a></li>
				22	<li><a href="#OtherItfs">Other useful <tt>AliasAnalysis</tt> methods</a></li>
				23	</ul>
				24	</li>
				25
				26	<li><a href="#writingnew">Writing a new <tt>AliasAnalysis</tt> Implementation</a>
				27	<ul>
				28	<li><a href="#passsubclasses">Different Pass styles</a></li>
				29	<li><a href="#requiredcalls">Required initialization calls</a></li>
				30	<li><a href="#interfaces">Interfaces which may be specified</a></li>
				31	<li><a href="#chaining"><tt>AliasAnalysis</tt> chaining behavior</a></li>
				32	<li><a href="#updating">Updating analysis results for transformations</a></li>
				33	<li><a href="#implefficiency">Efficiency Issues</a></li>
Dan Gohman	aa78544	2010-06-24 19:34:03 +0000	[diff] [blame]	34	<li><a href="#passmanager">Pass Manager Issues</a></li>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	35	</ul>
				36	</li>
				37
				38	<li><a href="#using">Using alias analysis results</a>
				39	<ul>
Chris Lattner	72da576	2009-04-25 21:11:37 +0000	[diff] [blame]	40	<li><a href="#memdep">Using the <tt>MemoryDependenceAnalysis</tt> Pass</a></li>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	41	<li><a href="#ast">Using the <tt>AliasSetTracker</tt> class</a></li>
				42	<li><a href="#direct">Using the <tt>AliasAnalysis</tt> interface directly</a></li>
				43	</ul>
				44	</li>
				45
				46	<li><a href="#exist">Existing alias analysis implementations and clients</a>
				47	<ul>
				48	<li><a href="#impls">Available <tt>AliasAnalysis</tt> implementations</a></li>
				49	<li><a href="#aliasanalysis-xforms">Alias analysis driven transformations</a></li>
				50	<li><a href="#aliasanalysis-debug">Clients for debugging and evaluation of
				51	implementations</a></li>
				52	</ul>
				53	</li>
Owen Anderson	4f10b31	2007-10-02 00:44:20 +0000	[diff] [blame]	54	<li><a href="#memdep">Memory Dependence Analysis</a></li>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	55	</ol>
				56
				57	<div class="doc_author">
				58	<p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p>
				59	</div>
				60
				61	<!-- *********************************************************************** -->
				62	<div class="doc_section">
				63	<a name="introduction">Introduction</a>
				64	</div>
				65	<!-- *********************************************************************** -->
				66
				67	<div class="doc_text">
				68
				69	<p>Alias Analysis (aka Pointer Analysis) is a class of techniques which attempt
				70	to determine whether or not two pointers ever can point to the same object in
				71	memory. There are many different algorithms for alias analysis and many
				72	different ways of classifying them: flow-sensitive vs flow-insensitive,
				73	context-sensitive vs context-insensitive, field-sensitive vs field-insensitive,
				74	unification-based vs subset-based, etc. Traditionally, alias analyses respond
				75	to a query with a <a href="#MustMayNo">Must, May, or No</a> alias response,
				76	indicating that two pointers always point to the same object, might point to the
				77	same object, or are known to never point to the same object.</p>
				78
				79	<p>The LLVM <a
				80	href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html"><tt>AliasAnalysis</tt></a>
				81	class is the primary interface used by clients and implementations of alias
				82	analyses in the LLVM system. This class is the common interface between clients
				83	of alias analysis information and the implementations providing it, and is
				84	designed to support a wide range of implementations and clients (but currently
				85	all clients are assumed to be flow-insensitive). In addition to simple alias
				86	analysis information, this class exposes Mod/Ref information from those
				87	implementations which can provide it, allowing for powerful analyses and
				88	transformations to work well together.</p>
				89
				90	<p>This document contains information necessary to successfully implement this
				91	interface, use it, and to test both sides. It also explains some of the finer
				92	points about what exactly results mean. If you feel that something is unclear
				93	or should be added, please <a href="mailto:sabre@nondot.org">let me
				94	know</a>.</p>
				95
				96	</div>
				97
				98	<!-- *********************************************************************** -->
				99	<div class="doc_section">
				100	<a name="overview"><tt>AliasAnalysis</tt> Class Overview</a>
				101	</div>
				102	<!-- *********************************************************************** -->
				103
				104	<div class="doc_text">
				105
				106	<p>The <a
				107	href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html"><tt>AliasAnalysis</tt></a>
				108	class defines the interface that the various alias analysis implementations
				109	should support. This class exports two important enums: <tt>AliasResult</tt>
				110	and <tt>ModRefResult</tt> which represent the result of an alias query or a
				111	mod/ref query, respectively.</p>
				112
				113	<p>The <tt>AliasAnalysis</tt> interface exposes information about memory,
				114	represented in several different ways. In particular, memory objects are
				115	represented as a starting address and size, and function calls are represented
				116	as the actual <tt>call</tt> or <tt>invoke</tt> instructions that performs the
				117	call. The <tt>AliasAnalysis</tt> interface also exposes some helper methods
				118	which allow you to get mod/ref information for arbitrary instructions.</p>
				119
				120	</div>
				121
				122	<!-- ======================================================================= -->
				123	<div class="doc_subsection">
				124	<a name="pointers">Representation of Pointers</a>
				125	</div>
				126
				127	<div class="doc_text">
				128
				129	<p>Most importantly, the <tt>AliasAnalysis</tt> class provides several methods
				130	which are used to query whether or not two memory objects alias, whether
				131	function calls can modify or read a memory object, etc. For all of these
				132	queries, memory objects are represented as a pair of their starting address (a
				133	symbolic LLVM <tt>Value*</tt>) and a static size.</p>
				134
				135	<p>Representing memory objects as a starting address and a size is critically
				136	important for correct Alias Analyses. For example, consider this (silly, but
				137	possible) C code:</p>
				138
				139	<div class="doc_code">
				140	<pre>
				141	int i;
				142	char C[2];
				143	char A[10];
				144	/* ... */
				145	for (i = 0; i != 10; ++i) {
				146	C[0] = A[i]; /* One byte store */
				147	C[1] = A[9-i]; /* One byte store */
				148	}
				149	</pre>
				150	</div>
				151
				152	<p>In this case, the <tt>basicaa</tt> pass will disambiguate the stores to
				153	<tt>C[0]</tt> and <tt>C[1]</tt> because they are accesses to two distinct
				154	locations one byte apart, and the accesses are each one byte. In this case, the
				155	LICM pass can use store motion to remove the stores from the loop. In
				156	constrast, the following code:</p>
				157
				158	<div class="doc_code">
				159	<pre>
				160	int i;
				161	char C[2];
				162	char A[10];
				163	/* ... */
				164	for (i = 0; i != 10; ++i) {
				165	((short)C)[0] = A[i]; / Two byte store! */
				166	C[1] = A[9-i]; /* One byte store */
				167	}
				168	</pre>
				169	</div>
				170
				171	<p>In this case, the two stores to C do alias each other, because the access to
				172	the <tt>&C[0]</tt> element is a two byte access. If size information wasn't
				173	available in the query, even the first case would have to conservatively assume
				174	that the accesses alias.</p>
				175
				176	</div>
				177
				178	<!-- ======================================================================= -->
				179	<div class="doc_subsection">
				180	<a name="alias">The <tt>alias</tt> method</a>
				181	</div>
				182
				183	<div class="doc_text">
				184	The <tt>alias</tt> method is the primary interface used to determine whether or
				185	not two memory objects alias each other. It takes two memory objects as input
				186	and returns MustAlias, MayAlias, or NoAlias as appropriate.
				187	</div>
				188
				189	<!-- _______________________________________________________________________ -->
				190	<div class="doc_subsubsection">
				191	<a name="MustMayNo">Must, May, and No Alias Responses</a>
				192	</div>
				193
				194	<div class="doc_text">
Dan Gohman	c820844	2010-07-02 18:41:32 +0000	[diff] [blame]	195	<p>The NoAlias response may be used when there is never an immediate dependence
				196	between any memory reference <i>based</i> on one pointer and any memory
				197	reference <i>based</i> the other. The most obvious example is when the two
				198	pointers point to non-overlapping memory ranges. Another is when the two
				199	pointers are only ever used for reading memory. Another is when the memory is
				200	freed and reallocated between accesses through one pointer and accesses through
				201	the other -- in this case, there is a dependence, but it's mediated by the free
				202	and reallocation.</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	203
Dan Gohman	24fc36d	2010-07-02 23:46:54 +0000	[diff] [blame^]	204	<p>As an exception to this is with the
				205	<a href="LangRef.html#noalias"><tt>noalias</tt></a> keyword. AliasAnalysis
				206	implementations may choose to ignore the "irrelevant" dependencies, provided
				207	their clients do not need to be aware of these dependencies for correctness.</p>
				208
Nick Lewycky	74c8756	2008-12-14 21:08:48 +0000	[diff] [blame]	209	<p>The MayAlias response is used whenever the two pointers might refer to the
				210	same object. If the two memory objects overlap, but do not start at the same
				211	location, return MayAlias.</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	212
Nick Lewycky	74c8756	2008-12-14 21:08:48 +0000	[diff] [blame]	213	<p>The MustAlias response may only be returned if the two memory objects are
				214	guaranteed to always start at exactly the same location. A MustAlias response
				215	implies that the pointers compare equal.</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	216
				217	</div>
				218
				219	<!-- ======================================================================= -->
				220	<div class="doc_subsection">
				221	<a name="ModRefInfo">The <tt>getModRefInfo</tt> methods</a>
				222	</div>
				223
				224	<div class="doc_text">
				225
				226	<p>The <tt>getModRefInfo</tt> methods return information about whether the
				227	execution of an instruction can read or modify a memory location. Mod/Ref
				228	information is always conservative: if an instruction <b>might</b> read or write
				229	a location, ModRef is returned.</p>
				230
				231	<p>The <tt>AliasAnalysis</tt> class also provides a <tt>getModRefInfo</tt>
				232	method for testing dependencies between function calls. This method takes two
				233	call sites (CS1 & CS2), returns NoModRef if the two calls refer to disjoint
				234	memory locations, Ref if CS1 reads memory written by CS2, Mod if CS1 writes to
				235	memory read or written by CS2, or ModRef if CS1 might read or write memory
Chris Lattner	3b80375	2009-11-22 16:01:44 +0000	[diff] [blame]	236	accessed by CS2. Note that this relation is not commutative.</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	237
				238	</div>
				239
				240
				241	<!-- ======================================================================= -->
				242	<div class="doc_subsection">
				243	<a name="OtherItfs">Other useful <tt>AliasAnalysis</tt> methods</a>
				244	</div>
				245
				246	<div class="doc_text">
				247
				248	<p>
				249	Several other tidbits of information are often collected by various alias
				250	analysis implementations and can be put to good use by various clients.
				251	</p>
				252
				253	</div>
				254
				255	<!-- _______________________________________________________________________ -->
				256	<div class="doc_subsubsection">
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	257	The <tt>pointsToConstantMemory</tt> method
				258	</div>
				259
				260	<div class="doc_text">
				261
				262	<p>The <tt>pointsToConstantMemory</tt> method returns true if and only if the
				263	analysis can prove that the pointer only points to unchanging memory locations
				264	(functions, constant global variables, and the null pointer). This information
				265	can be used to refine mod/ref information: it is impossible for an unchanging
				266	memory location to be modified.</p>
				267
				268	</div>
				269
				270	<!-- _______________________________________________________________________ -->
				271	<div class="doc_subsubsection">
				272	<a name="simplemodref">The <tt>doesNotAccessMemory</tt> and
				273	<tt>onlyReadsMemory</tt> methods</a>
				274	</div>
				275
				276	<div class="doc_text">
				277
				278	<p>These methods are used to provide very simple mod/ref information for
				279	function calls. The <tt>doesNotAccessMemory</tt> method returns true for a
				280	function if the analysis can prove that the function never reads or writes to
				281	memory, or if the function only reads from constant memory. Functions with this
				282	property are side-effect free and only depend on their input arguments, allowing
				283	them to be eliminated if they form common subexpressions or be hoisted out of
				284	loops. Many common functions behave this way (e.g., <tt>sin</tt> and
				285	<tt>cos</tt>) but many others do not (e.g., <tt>acos</tt>, which modifies the
				286	<tt>errno</tt> variable).</p>
				287
				288	<p>The <tt>onlyReadsMemory</tt> method returns true for a function if analysis
				289	can prove that (at most) the function only reads from non-volatile memory.
				290	Functions with this property are side-effect free, only depending on their input
				291	arguments and the state of memory when they are called. This property allows
				292	calls to these functions to be eliminated and moved around, as long as there is
				293	no store instruction that changes the contents of memory. Note that all
				294	functions that satisfy the <tt>doesNotAccessMemory</tt> method also satisfies
				295	<tt>onlyReadsMemory</tt>.</p>
				296
				297	</div>
				298
				299	<!-- *********************************************************************** -->
				300	<div class="doc_section">
				301	<a name="writingnew">Writing a new <tt>AliasAnalysis</tt> Implementation</a>
				302	</div>
				303	<!-- *********************************************************************** -->
				304
				305	<div class="doc_text">
				306
				307	<p>Writing a new alias analysis implementation for LLVM is quite
				308	straight-forward. There are already several implementations that you can use
				309	for examples, and the following information should help fill in any details.
				310	For a examples, take a look at the <a href="#impls">various alias analysis
				311	implementations</a> included with LLVM.</p>
				312
				313	</div>
				314
				315	<!-- ======================================================================= -->
				316	<div class="doc_subsection">
				317	<a name="passsubclasses">Different Pass styles</a>
				318	</div>
				319
				320	<div class="doc_text">
				321
				322	<p>The first step to determining what type of <a
				323	href="WritingAnLLVMPass.html">LLVM pass</a> you need to use for your Alias
				324	Analysis. As is the case with most other analyses and transformations, the
				325	answer should be fairly obvious from what type of problem you are trying to
				326	solve:</p>
				327
				328	<ol>
				329	<li>If you require interprocedural analysis, it should be a
				330	<tt>Pass</tt>.</li>
				331	<li>If you are a function-local analysis, subclass <tt>FunctionPass</tt>.</li>
				332	<li>If you don't need to look at the program at all, subclass
				333	<tt>ImmutablePass</tt>.</li>
				334	</ol>
				335
				336	<p>In addition to the pass that you subclass, you should also inherit from the
				337	<tt>AliasAnalysis</tt> interface, of course, and use the
				338	<tt>RegisterAnalysisGroup</tt> template to register as an implementation of
				339	<tt>AliasAnalysis</tt>.</p>
				340
				341	</div>
				342
				343	<!-- ======================================================================= -->
				344	<div class="doc_subsection">
				345	<a name="requiredcalls">Required initialization calls</a>
				346	</div>
				347
				348	<div class="doc_text">
				349
				350	<p>Your subclass of <tt>AliasAnalysis</tt> is required to invoke two methods on
				351	the <tt>AliasAnalysis</tt> base class: <tt>getAnalysisUsage</tt> and
				352	<tt>InitializeAliasAnalysis</tt>. In particular, your implementation of
				353	<tt>getAnalysisUsage</tt> should explicitly call into the
				354	<tt>AliasAnalysis::getAnalysisUsage</tt> method in addition to doing any
				355	declaring any pass dependencies your pass has. Thus you should have something
				356	like this:</p>
				357
				358	<div class="doc_code">
				359	<pre>
				360	void getAnalysisUsage(AnalysisUsage &AU) const {
				361	AliasAnalysis::getAnalysisUsage(AU);
				362	<i>// declare your dependencies here.</i>
				363	}
				364	</pre>
				365	</div>
				366
				367	<p>Additionally, your must invoke the <tt>InitializeAliasAnalysis</tt> method
				368	from your analysis run method (<tt>run</tt> for a <tt>Pass</tt>,
				369	<tt>runOnFunction</tt> for a <tt>FunctionPass</tt>, or <tt>InitializePass</tt>
				370	for an <tt>ImmutablePass</tt>). For example (as part of a <tt>Pass</tt>):</p>
				371
				372	<div class="doc_code">
				373	<pre>
				374	bool run(Module &M) {
				375	InitializeAliasAnalysis(this);
				376	<i>// Perform analysis here...</i>
				377	return false;
				378	}
				379	</pre>
				380	</div>
				381
				382	</div>
				383
				384	<!-- ======================================================================= -->
				385	<div class="doc_subsection">
				386	<a name="interfaces">Interfaces which may be specified</a>
				387	</div>
				388
				389	<div class="doc_text">
				390
				391	<p>All of the <a
				392	href="/doxygen/classllvm_1_1AliasAnalysis.html"><tt>AliasAnalysis</tt></a>
				393	virtual methods default to providing <a href="#chaining">chaining</a> to another
				394	alias analysis implementation, which ends up returning conservatively correct
				395	information (returning "May" Alias and "Mod/Ref" for alias and mod/ref queries
				396	respectively). Depending on the capabilities of the analysis you are
				397	implementing, you just override the interfaces you can improve.</p>
				398
				399	</div>
				400
				401
				402
				403	<!-- ======================================================================= -->
				404	<div class="doc_subsection">
				405	<a name="chaining"><tt>AliasAnalysis</tt> chaining behavior</a>
				406	</div>
				407
				408	<div class="doc_text">
				409
				410	<p>With only two special exceptions (the <tt><a
				411	href="#basic-aa">basicaa</a></tt> and <a href="#no-aa"><tt>no-aa</tt></a>
				412	passes) every alias analysis pass chains to another alias analysis
				413	implementation (for example, the user can specify "<tt>-basicaa -ds-aa
Chris Lattner	5dd9205	2010-03-01 19:24:17 +0000	[diff] [blame]	414	-licm</tt>" to get the maximum benefit from both alias
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	415	analyses). The alias analysis class automatically takes care of most of this
				416	for methods that you don't override. For methods that you do override, in code
				417	paths that return a conservative MayAlias or Mod/Ref result, simply return
				418	whatever the superclass computes. For example:</p>
				419
				420	<div class="doc_code">
				421	<pre>
				422	AliasAnalysis::AliasResult alias(const Value *V1, unsigned V1Size,
				423	const Value *V2, unsigned V2Size) {
				424	if (...)
				425	return NoAlias;
				426	...
				427
				428	<i>// Couldn't determine a must or no-alias result.</i>
				429	return AliasAnalysis::alias(V1, V1Size, V2, V2Size);
				430	}
				431	</pre>
				432	</div>
				433
				434	<p>In addition to analysis queries, you must make sure to unconditionally pass
				435	LLVM <a href="#updating">update notification</a> methods to the superclass as
				436	well if you override them, which allows all alias analyses in a change to be
				437	updated.</p>
				438
				439	</div>
				440
				441
				442	<!-- ======================================================================= -->
				443	<div class="doc_subsection">
				444	<a name="updating">Updating analysis results for transformations</a>
				445	</div>
				446
				447	<div class="doc_text">
				448	<p>
				449	Alias analysis information is initially computed for a static snapshot of the
				450	program, but clients will use this information to make transformations to the
				451	code. All but the most trivial forms of alias analysis will need to have their
				452	analysis results updated to reflect the changes made by these transformations.
				453	</p>
				454
				455	<p>
				456	The <tt>AliasAnalysis</tt> interface exposes two methods which are used to
				457	communicate program changes from the clients to the analysis implementations.
				458	Various alias analysis implementations should use these methods to ensure that
				459	their internal data structures are kept up-to-date as the program changes (for
				460	example, when an instruction is deleted), and clients of alias analysis must be
				461	sure to call these interfaces appropriately.
				462	</p>
				463	</div>
				464
				465	<!-- _______________________________________________________________________ -->
				466	<div class="doc_subsubsection">The <tt>deleteValue</tt> method</div>
				467
				468	<div class="doc_text">
				469	The <tt>deleteValue</tt> method is called by transformations when they remove an
				470	instruction or any other value from the program (including values that do not
				471	use pointers). Typically alias analyses keep data structures that have entries
				472	for each value in the program. When this method is called, they should remove
				473	any entries for the specified value, if they exist.
				474	</div>
				475
				476	<!-- _______________________________________________________________________ -->
				477	<div class="doc_subsubsection">The <tt>copyValue</tt> method</div>
				478
				479	<div class="doc_text">
				480	The <tt>copyValue</tt> method is used when a new value is introduced into the
				481	program. There is no way to introduce a value into the program that did not
				482	exist before (this doesn't make sense for a safe compiler transformation), so
				483	this is the only way to introduce a new value. This method indicates that the
				484	new value has exactly the same properties as the value being copied.
				485	</div>
				486
				487	<!-- _______________________________________________________________________ -->
				488	<div class="doc_subsubsection">The <tt>replaceWithNewValue</tt> method</div>
				489
				490	<div class="doc_text">
				491	This method is a simple helper method that is provided to make clients easier to
				492	use. It is implemented by copying the old analysis information to the new
				493	value, then deleting the old value. This method cannot be overridden by alias
				494	analysis implementations.
				495	</div>
				496
				497	<!-- ======================================================================= -->
				498	<div class="doc_subsection">
				499	<a name="implefficiency">Efficiency Issues</a>
				500	</div>
				501
				502	<div class="doc_text">
				503
				504	<p>From the LLVM perspective, the only thing you need to do to provide an
				505	efficient alias analysis is to make sure that alias analysis <b>queries</b> are
				506	serviced quickly. The actual calculation of the alias analysis results (the
				507	"run" method) is only performed once, but many (perhaps duplicate) queries may
				508	be performed. Because of this, try to move as much computation to the run
				509	method as possible (within reason).</p>
				510
				511	</div>
				512
Dan Gohman	aa78544	2010-06-24 19:34:03 +0000	[diff] [blame]	513	<!-- ======================================================================= -->
				514	<div class="doc_subsection">
				515	<a name="passmanager">Pass Manager Issues</a>
				516	</div>
				517
				518	<div class="doc_text">
				519
				520	<p>PassManager support for alternative AliasAnalysis implementation
				521	has some issues.</p>
				522
				523	<p>There is no way to override the default alias analysis. It would
				524	be very useful to be able to do something like "opt -my-aa -O2" and
				525	have it use -my-aa for all passes which need AliasAnalysis, but there
				526	is currently no support for that, short of changing the source code
				527	and recompiling. Similarly, there is also no way of setting a chain
				528	of analyses as the default.</p>
				529
				530	<p>There is no way for transform passes to declare that they preserve
				531	<tt>AliasAnalysis</tt> implementations. The <tt>AliasAnalysis</tt>
				532	interface includes <tt>deleteValue</tt> and <tt>copyValue</tt> methods
				533	which are intended to allow a pass to keep an AliasAnalysis consistent,
				534	however there's no way for a pass to declare in its
				535	<tt>getAnalysisUsage</tt> that it does so. Some passes attempt to use
				536	<tt>AU.addPreserved<AliasAnalysis></tt>, however this doesn't
				537	actually have any effect.</tt>
				538
				539	<p><tt>AliasAnalysisCounter</tt> (<tt>-count-aa</tt>) and <tt>AliasDebugger</tt>
				540	(<tt>-debug-aa</tt>) are implemented as <tt>ModulePass</tt> classes, so if your
				541	alias analysis uses <tt>FunctionPass</tt>, it won't be able to use
				542	these utilities. If you try to use them, the pass manager will
				543	silently route alias analysis queries directly to
				544	<tt>BasicAliasAnalysis</tt> instead.</p>
				545
				546	<p>Similarly, the <tt>opt -p</tt> option introduces <tt>ModulePass</tt>
				547	passes between each pass, which prevents the use of <tt>FunctionPass</tt>
				548	alias analysis passes.</p>
				549
				550	</div>
				551
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	552	<!-- *********************************************************************** -->
				553	<div class="doc_section">
				554	<a name="using">Using alias analysis results</a>
				555	</div>
				556	<!-- *********************************************************************** -->
				557
				558	<div class="doc_text">
				559
				560	<p>There are several different ways to use alias analysis results. In order of
				561	preference, these are...</p>
				562
				563	</div>
				564
				565	<!-- ======================================================================= -->
				566	<div class="doc_subsection">
Chris Lattner	72da576	2009-04-25 21:11:37 +0000	[diff] [blame]	567	<a name="memdep">Using the <tt>MemoryDependenceAnalysis</tt> Pass</a>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	568	</div>
				569
				570	<div class="doc_text">
				571
Chris Lattner	72da576	2009-04-25 21:11:37 +0000	[diff] [blame]	572	<p>The <tt>memdep</tt> pass uses alias analysis to provide high-level dependence
				573	information about memory-using instructions. This will tell you which store
				574	feeds into a load, for example. It uses caching and other techniques to be
				575	efficient, and is used by Dead Store Elimination, GVN, and memcpy optimizations.
				576	</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	577
				578	</div>
				579
				580	<!-- ======================================================================= -->
				581	<div class="doc_subsection">
				582	<a name="ast">Using the <tt>AliasSetTracker</tt> class</a>
				583	</div>
				584
				585	<div class="doc_text">
				586
				587	<p>Many transformations need information about alias <b>sets</b> that are active
				588	in some scope, rather than information about pairwise aliasing. The <tt><a
				589	href="/doxygen/classllvm_1_1AliasSetTracker.html">AliasSetTracker</a></tt> class
				590	is used to efficiently build these Alias Sets from the pairwise alias analysis
				591	information provided by the <tt>AliasAnalysis</tt> interface.</p>
				592
				593	<p>First you initialize the AliasSetTracker by using the "<tt>add</tt>" methods
				594	to add information about various potentially aliasing instructions in the scope
				595	you are interested in. Once all of the alias sets are completed, your pass
				596	should simply iterate through the constructed alias sets, using the
				597	<tt>AliasSetTracker</tt> <tt>begin()</tt>/<tt>end()</tt> methods.</p>
				598
				599	<p>The <tt>AliasSet</tt>s formed by the <tt>AliasSetTracker</tt> are guaranteed
				600	to be disjoint, calculate mod/ref information and volatility for the set, and
				601	keep track of whether or not all of the pointers in the set are Must aliases.
				602	The AliasSetTracker also makes sure that sets are properly folded due to call
				603	instructions, and can provide a list of pointers in each set.</p>
				604
				605	<p>As an example user of this, the <a href="/doxygen/structLICM.html">Loop
				606	Invariant Code Motion</a> pass uses <tt>AliasSetTracker</tt>s to calculate alias
				607	sets for each loop nest. If an <tt>AliasSet</tt> in a loop is not modified,
				608	then all load instructions from that set may be hoisted out of the loop. If any
				609	alias sets are stored to <b>and</b> are must alias sets, then the stores may be
				610	sunk to outside of the loop, promoting the memory location to a register for the
				611	duration of the loop nest. Both of these transformations only apply if the
				612	pointer argument is loop-invariant.</p>
				613
				614	</div>
				615
				616	<!-- _______________________________________________________________________ -->
				617	<div class="doc_subsubsection">
				618	The AliasSetTracker implementation
				619	</div>
				620
				621	<div class="doc_text">
				622
				623	<p>The AliasSetTracker class is implemented to be as efficient as possible. It
				624	uses the union-find algorithm to efficiently merge AliasSets when a pointer is
				625	inserted into the AliasSetTracker that aliases multiple sets. The primary data
				626	structure is a hash table mapping pointers to the AliasSet they are in.</p>
				627
				628	<p>The AliasSetTracker class must maintain a list of all of the LLVM Value*'s
				629	that are in each AliasSet. Since the hash table already has entries for each
				630	LLVM Value* of interest, the AliasesSets thread the linked list through these
				631	hash-table nodes to avoid having to allocate memory unnecessarily, and to make
				632	merging alias sets extremely efficient (the linked list merge is constant time).
				633	</p>
				634
				635	<p>You shouldn't need to understand these details if you are just a client of
				636	the AliasSetTracker, but if you look at the code, hopefully this brief
				637	description will help make sense of why things are designed the way they
				638	are.</p>
				639
				640	</div>
				641
				642	<!-- ======================================================================= -->
				643	<div class="doc_subsection">
				644	<a name="direct">Using the <tt>AliasAnalysis</tt> interface directly</a>
				645	</div>
				646
				647	<div class="doc_text">
				648
				649	<p>If neither of these utility class are what your pass needs, you should use
				650	the interfaces exposed by the <tt>AliasAnalysis</tt> class directly. Try to use
				651	the higher-level methods when possible (e.g., use mod/ref information instead of
				652	the <a href="#alias"><tt>alias</tt></a> method directly if possible) to get the
				653	best precision and efficiency.</p>
				654
				655	</div>
				656
				657	<!-- *********************************************************************** -->
				658	<div class="doc_section">
				659	<a name="exist">Existing alias analysis implementations and clients</a>
				660	</div>
				661	<!-- *********************************************************************** -->
				662
				663	<div class="doc_text">
				664
				665	<p>If you're going to be working with the LLVM alias analysis infrastructure,
				666	you should know what clients and implementations of alias analysis are
				667	available. In particular, if you are implementing an alias analysis, you should
				668	be aware of the <a href="#aliasanalysis-debug">the clients</a> that are useful
				669	for monitoring and evaluating different implementations.</p>
				670
				671	</div>
				672
				673	<!-- ======================================================================= -->
				674	<div class="doc_subsection">
				675	<a name="impls">Available <tt>AliasAnalysis</tt> implementations</a>
				676	</div>
				677
				678	<div class="doc_text">
				679
				680	<p>This section lists the various implementations of the <tt>AliasAnalysis</tt>
				681	interface. With the exception of the <a href="#no-aa"><tt>-no-aa</tt></a> and
				682	<a href="#basic-aa"><tt>-basicaa</tt></a> implementations, all of these <a
				683	href="#chaining">chain</a> to other alias analysis implementations.</p>
				684
				685	</div>
				686
				687	<!-- _______________________________________________________________________ -->
				688	<div class="doc_subsubsection">
				689	<a name="no-aa">The <tt>-no-aa</tt> pass</a>
				690	</div>
				691
				692	<div class="doc_text">
				693
				694	<p>The <tt>-no-aa</tt> pass is just like what it sounds: an alias analysis that
				695	never returns any useful information. This pass can be useful if you think that
				696	alias analysis is doing something wrong and are trying to narrow down a
				697	problem.</p>
				698
				699	</div>
				700
				701	<!-- _______________________________________________________________________ -->
				702	<div class="doc_subsubsection">
				703	<a name="basic-aa">The <tt>-basicaa</tt> pass</a>
				704	</div>
				705
				706	<div class="doc_text">
				707
				708	<p>The <tt>-basicaa</tt> pass is the default LLVM alias analysis. It is an
				709	aggressive local analysis that "knows" many important facts:</p>
				710
				711	<ul>
				712	<li>Distinct globals, stack allocations, and heap allocations can never
				713	alias.</li>
				714	<li>Globals, stack allocations, and heap allocations never alias the null
				715	pointer.</li>
				716	<li>Different fields of a structure do not alias.</li>
				717	<li>Indexes into arrays with statically differing subscripts cannot alias.</li>
				718	<li>Many common standard C library functions <a
				719	href="#simplemodref">never access memory or only read memory</a>.</li>
				720	<li>Pointers that obviously point to constant globals
				721	"<tt>pointToConstantMemory</tt>".</li>
				722	<li>Function calls can not modify or references stack allocations if they never
				723	escape from the function that allocates them (a common case for automatic
				724	arrays).</li>
				725	</ul>
				726
				727	</div>
				728
				729	<!-- _______________________________________________________________________ -->
				730	<div class="doc_subsubsection">
				731	<a name="globalsmodref">The <tt>-globalsmodref-aa</tt> pass</a>
				732	</div>
				733
				734	<div class="doc_text">
				735
				736	<p>This pass implements a simple context-sensitive mod/ref and alias analysis
				737	for internal global variables that don't "have their address taken". If a
				738	global does not have its address taken, the pass knows that no pointers alias
				739	the global. This pass also keeps track of functions that it knows never access
Chris Lattner	72da576	2009-04-25 21:11:37 +0000	[diff] [blame]	740	memory or never read memory. This allows certain optimizations (e.g. GVN) to
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	741	eliminate call instructions entirely.
				742	</p>
				743
				744	<p>The real power of this pass is that it provides context-sensitive mod/ref
				745	information for call instructions. This allows the optimizer to know that
				746	calls to a function do not clobber or read the value of the global, allowing
				747	loads and stores to be eliminated.</p>
				748
				749	<p>Note that this pass is somewhat limited in its scope (only support
				750	non-address taken globals), but is very quick analysis.</p>
				751	</div>
				752
				753	<!-- _______________________________________________________________________ -->
				754	<div class="doc_subsubsection">
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	755	<a name="steens-aa">The <tt>-steens-aa</tt> pass</a>
				756	</div>
				757
				758	<div class="doc_text">
				759
				760	<p>The <tt>-steens-aa</tt> pass implements a variation on the well-known
				761	"Steensgaard's algorithm" for interprocedural alias analysis. Steensgaard's
				762	algorithm is a unification-based, flow-insensitive, context-insensitive, and
				763	field-insensitive alias analysis that is also very scalable (effectively linear
				764	time).</p>
				765
				766	<p>The LLVM <tt>-steens-aa</tt> pass implements a "speculatively
				767	field-<b>sensitive</b>" version of Steensgaard's algorithm using the Data
				768	Structure Analysis framework. This gives it substantially more precision than
				769	the standard algorithm while maintaining excellent analysis scalability.</p>
				770
				771	<p>Note that <tt>-steens-aa</tt> is available in the optional "poolalloc"
				772	module, it is not part of the LLVM core.</p>
				773
				774	</div>
				775
				776	<!-- _______________________________________________________________________ -->
				777	<div class="doc_subsubsection">
				778	<a name="ds-aa">The <tt>-ds-aa</tt> pass</a>
				779	</div>
				780
				781	<div class="doc_text">
				782
				783	<p>The <tt>-ds-aa</tt> pass implements the full Data Structure Analysis
				784	algorithm. Data Structure Analysis is a modular unification-based,
				785	flow-insensitive, context-<b>sensitive</b>, and speculatively
				786	field-<b>sensitive</b> alias analysis that is also quite scalable, usually at
				787	O(n*log(n)).</p>
				788
				789	<p>This algorithm is capable of responding to a full variety of alias analysis
				790	queries, and can provide context-sensitive mod/ref information as well. The
				791	only major facility not implemented so far is support for must-alias
				792	information.</p>
				793
				794	<p>Note that <tt>-ds-aa</tt> is available in the optional "poolalloc"
				795	module, it is not part of the LLVM core.</p>
				796
				797	</div>
				798
Dan Gohman	3e6ba36	2010-06-28 22:09:52 +0000	[diff] [blame]	799	<!-- _______________________________________________________________________ -->
				800	<div class="doc_subsubsection">
				801	<a name="scev-aa">The <tt>-scev-aa</tt> pass</a>
				802	</div>
				803
				804	<div class="doc_text">
				805
				806	<p>The <tt>-scev-aa</tt> pass implements AliasAnalysis queries by
				807	translating them into ScalarEvolution queries. This gives it a
				808	more complete understanding of <tt>getelementptr</tt> instructions
				809	and loop induction variables than other alias analyses have.</p>
				810
				811	</div>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	812
				813	<!-- ======================================================================= -->
				814	<div class="doc_subsection">
				815	<a name="aliasanalysis-xforms">Alias analysis driven transformations</a>
				816	</div>
				817
				818	<div class="doc_text">
				819	LLVM includes several alias-analysis driven transformations which can be used
				820	with any of the implementations above.
				821	</div>
				822
				823	<!-- _______________________________________________________________________ -->
				824	<div class="doc_subsubsection">
				825	<a name="adce">The <tt>-adce</tt> pass</a>
				826	</div>
				827
				828	<div class="doc_text">
				829
				830	<p>The <tt>-adce</tt> pass, which implements Aggressive Dead Code Elimination
				831	uses the <tt>AliasAnalysis</tt> interface to delete calls to functions that do
				832	not have side-effects and are not used.</p>
				833
				834	</div>
				835
				836
				837	<!-- _______________________________________________________________________ -->
				838	<div class="doc_subsubsection">
				839	<a name="licm">The <tt>-licm</tt> pass</a>
				840	</div>
				841
				842	<div class="doc_text">
				843
				844	<p>The <tt>-licm</tt> pass implements various Loop Invariant Code Motion related
				845	transformations. It uses the <tt>AliasAnalysis</tt> interface for several
				846	different transformations:</p>
				847
				848	<ul>
				849	<li>It uses mod/ref information to hoist or sink load instructions out of loops
				850	if there are no instructions in the loop that modifies the memory loaded.</li>
				851
				852	<li>It uses mod/ref information to hoist function calls out of loops that do not
				853	write to memory and are loop-invariant.</li>
				854
				855	<li>If uses alias information to promote memory objects that are loaded and
				856	stored to in loops to live in a register instead. It can do this if there are
				857	no may aliases to the loaded/stored memory location.</li>
				858	</ul>
				859
				860	</div>
				861
				862	<!-- _______________________________________________________________________ -->
				863	<div class="doc_subsubsection">
				864	<a name="argpromotion">The <tt>-argpromotion</tt> pass</a>
				865	</div>
				866
				867	<div class="doc_text">
				868	<p>
				869	The <tt>-argpromotion</tt> pass promotes by-reference arguments to be passed in
				870	by-value instead. In particular, if pointer arguments are only loaded from it
				871	passes in the value loaded instead of the address to the function. This pass
				872	uses alias information to make sure that the value loaded from the argument
				873	pointer is not modified between the entry of the function and any load of the
				874	pointer.</p>
				875	</div>
				876
				877	<!-- _______________________________________________________________________ -->
				878	<div class="doc_subsubsection">
Chris Lattner	72da576	2009-04-25 21:11:37 +0000	[diff] [blame]	879	<a name="gvn">The <tt>-gvn</tt>, <tt>-memcpyopt</tt>, and <tt>-dse</tt>
				880	passes</a>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	881	</div>
				882
				883	<div class="doc_text">
				884
Chris Lattner	72da576	2009-04-25 21:11:37 +0000	[diff] [blame]	885	<p>These passes use AliasAnalysis information to reason about loads and stores.
				886	</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	887
				888	</div>
				889
				890	<!-- ======================================================================= -->
				891	<div class="doc_subsection">
				892	<a name="aliasanalysis-debug">Clients for debugging and evaluation of
				893	implementations</a>
				894	</div>
				895
				896	<div class="doc_text">
				897
				898	<p>These passes are useful for evaluating the various alias analysis
Chris Lattner	5dd9205	2010-03-01 19:24:17 +0000	[diff] [blame]	899	implementations. You can use them with commands like '<tt>opt -ds-aa
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	900	-aa-eval foo.bc -disable-output -stats</tt>'.</p>
				901
				902	</div>
				903
				904	<!-- _______________________________________________________________________ -->
				905	<div class="doc_subsubsection">
				906	<a name="print-alias-sets">The <tt>-print-alias-sets</tt> pass</a>
				907	</div>
				908
				909	<div class="doc_text">
				910
				911	<p>The <tt>-print-alias-sets</tt> pass is exposed as part of the
				912	<tt>opt</tt> tool to print out the Alias Sets formed by the <a
				913	href="#ast"><tt>AliasSetTracker</tt></a> class. This is useful if you're using
				914	the <tt>AliasSetTracker</tt> class. To use it, use something like:</p>
				915
				916	<div class="doc_code">
				917	<pre>
				918	% opt -ds-aa -print-alias-sets -disable-output
				919	</pre>
				920	</div>
				921
				922	</div>
				923
				924
				925	<!-- _______________________________________________________________________ -->
				926	<div class="doc_subsubsection">
				927	<a name="count-aa">The <tt>-count-aa</tt> pass</a>
				928	</div>
				929
				930	<div class="doc_text">
				931
				932	<p>The <tt>-count-aa</tt> pass is useful to see how many queries a particular
				933	pass is making and what responses are returned by the alias analysis. As an
				934	example,</p>
				935
				936	<div class="doc_code">
				937	<pre>
				938	% opt -basicaa -count-aa -ds-aa -count-aa -licm
				939	</pre>
				940	</div>
				941
				942	<p>will print out how many queries (and what responses are returned) by the
				943	<tt>-licm</tt> pass (of the <tt>-ds-aa</tt> pass) and how many queries are made
				944	of the <tt>-basicaa</tt> pass by the <tt>-ds-aa</tt> pass. This can be useful
				945	when debugging a transformation or an alias analysis implementation.</p>
				946
				947	</div>
				948
				949	<!-- _______________________________________________________________________ -->
				950	<div class="doc_subsubsection">
				951	<a name="aa-eval">The <tt>-aa-eval</tt> pass</a>
				952	</div>
				953
				954	<div class="doc_text">
				955
				956	<p>The <tt>-aa-eval</tt> pass simply iterates through all pairs of pointers in a
				957	function and asks an alias analysis whether or not the pointers alias. This
				958	gives an indication of the precision of the alias analysis. Statistics are
				959	printed indicating the percent of no/may/must aliases found (a more precise
				960	algorithm will have a lower number of may aliases).</p>
				961
				962	</div>
				963
				964	<!-- *********************************************************************** -->
Owen Anderson	05e080f	2007-10-02 00:43:25 +0000	[diff] [blame]	965	<div class="doc_section">
				966	<a name="memdep">Memory Dependence Analysis</a>
				967	</div>
				968	<!-- *********************************************************************** -->
				969
				970	<div class="doc_text">
				971
				972	<p>If you're just looking to be a client of alias analysis information, consider
				973	using the Memory Dependence Analysis interface instead. MemDep is a lazy,
				974	caching layer on top of alias analysis that is able to answer the question of
				975	what preceding memory operations a given instruction depends on, either at an
				976	intra- or inter-block level. Because of its laziness and caching
				977	policy, using MemDep can be a significant performance win over accessing alias
				978	analysis directly.</p>
				979
				980	</div>
				981
				982	<!-- *********************************************************************** -->
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	983
				984	<hr>
				985	<address>
				986	<a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman	947321d	2008-12-11 17:34:48 +0000	[diff] [blame]	987	src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	988	<a href="http://validator.w3.org/check/referer"><img
Misha Brukman	947321d	2008-12-11 17:34:48 +0000	[diff] [blame]	989	src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	990
				991	<a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
				992	<a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
				993	Last modified: $Date$
				994	</address>
				995
				996	</body>
				997	</html>