Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / llvm / 05e080fe9f0b5495a764c87c70467edaea8ff639 / . / docs / AliasAnalysis.html
blob: 7c3fa81d443af95208045478c18cadf7feeefb01 [file] [log] [blame]
Dan Gohmanf17a25c2007-07-18 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>
34 </ul>
35 </li>
36
37 <li><a href="#using">Using alias analysis results</a>
38 <ul>
39 <li><a href="#loadvn">Using the <tt>-load-vn</tt> Pass</a></li>
40 <li><a href="#ast">Using the <tt>AliasSetTracker</tt> class</a></li>
41 <li><a href="#direct">Using the <tt>AliasAnalysis</tt> interface directly</a></li>
42 </ul>
43 </li>
44
45 <li><a href="#exist">Existing alias analysis implementations and clients</a>
46 <ul>
47 <li><a href="#impls">Available <tt>AliasAnalysis</tt> implementations</a></li>
48 <li><a href="#aliasanalysis-xforms">Alias analysis driven transformations</a></li>
49 <li><a href="#aliasanalysis-debug">Clients for debugging and evaluation of
50 implementations</a></li>
51 </ul>
52 </li>
Owen Anderson05e080f2007-10-02 00:43:25 +0000[diff] [blame^]53 <li><a hread="#memdep">Memory Dependence Analysis</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]54</ol>
55
56<div class="doc_author">
57 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p>
58</div>
59
60<!-- *********************************************************************** -->
61<div class="doc_section">
62 <a name="introduction">Introduction</a>
63</div>
64<!-- *********************************************************************** -->
65
66<div class="doc_text">
67
68<p>Alias Analysis (aka Pointer Analysis) is a class of techniques which attempt
69to determine whether or not two pointers ever can point to the same object in
70memory. There are many different algorithms for alias analysis and many
71different ways of classifying them: flow-sensitive vs flow-insensitive,
72context-sensitive vs context-insensitive, field-sensitive vs field-insensitive,
73unification-based vs subset-based, etc. Traditionally, alias analyses respond
74to a query with a <a href="#MustMayNo">Must, May, or No</a> alias response,
75indicating that two pointers always point to the same object, might point to the
76same object, or are known to never point to the same object.</p>
77
78<p>The LLVM <a
79href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html"><tt>AliasAnalysis</tt></a>
80class is the primary interface used by clients and implementations of alias
81analyses in the LLVM system. This class is the common interface between clients
82of alias analysis information and the implementations providing it, and is
83designed to support a wide range of implementations and clients (but currently
84all clients are assumed to be flow-insensitive). In addition to simple alias
85analysis information, this class exposes Mod/Ref information from those
86implementations which can provide it, allowing for powerful analyses and
87transformations to work well together.</p>
88
89<p>This document contains information necessary to successfully implement this
90interface, use it, and to test both sides. It also explains some of the finer
91points about what exactly results mean. If you feel that something is unclear
92or should be added, please <a href="mailto:sabre@nondot.org">let me
93know</a>.</p>
94
95</div>
96
97<!-- *********************************************************************** -->
98<div class="doc_section">
99 <a name="overview"><tt>AliasAnalysis</tt> Class Overview</a>
100</div>
101<!-- *********************************************************************** -->
102
103<div class="doc_text">
104
105<p>The <a
106href="http://llvm.org/doxygen/classllvm_1_1AliasAnalysis.html"><tt>AliasAnalysis</tt></a>
107class defines the interface that the various alias analysis implementations
108should support. This class exports two important enums: <tt>AliasResult</tt>
109and <tt>ModRefResult</tt> which represent the result of an alias query or a
110mod/ref query, respectively.</p>
111
112<p>The <tt>AliasAnalysis</tt> interface exposes information about memory,
113represented in several different ways. In particular, memory objects are
114represented as a starting address and size, and function calls are represented
115as the actual <tt>call</tt> or <tt>invoke</tt> instructions that performs the
116call. The <tt>AliasAnalysis</tt> interface also exposes some helper methods
117which allow you to get mod/ref information for arbitrary instructions.</p>
118
119</div>
120
121<!-- ======================================================================= -->
122<div class="doc_subsection">
123 <a name="pointers">Representation of Pointers</a>
124</div>
125
126<div class="doc_text">
127
128<p>Most importantly, the <tt>AliasAnalysis</tt> class provides several methods
129which are used to query whether or not two memory objects alias, whether
130function calls can modify or read a memory object, etc. For all of these
131queries, memory objects are represented as a pair of their starting address (a
132symbolic LLVM <tt>Value*</tt>) and a static size.</p>
133
134<p>Representing memory objects as a starting address and a size is critically
135important for correct Alias Analyses. For example, consider this (silly, but
136possible) C code:</p>
137
138<div class="doc_code">
139<pre>
140int i;
141char C[2];
142char A[10];
143/* ... */
144for (i = 0; i != 10; ++i) {
145 C[0] = A[i]; /* One byte store */
146 C[1] = A[9-i]; /* One byte store */
147}
148</pre>
149</div>
150
151<p>In this case, the <tt>basicaa</tt> pass will disambiguate the stores to
152<tt>C[0]</tt> and <tt>C[1]</tt> because they are accesses to two distinct
153locations one byte apart, and the accesses are each one byte. In this case, the
154LICM pass can use store motion to remove the stores from the loop. In
155constrast, the following code:</p>
156
157<div class="doc_code">
158<pre>
159int i;
160char C[2];
161char A[10];
162/* ... */
163for (i = 0; i != 10; ++i) {
164 ((short*)C)[0] = A[i]; /* Two byte store! */
165 C[1] = A[9-i]; /* One byte store */
166}
167</pre>
168</div>
169
170<p>In this case, the two stores to C do alias each other, because the access to
171the <tt>&amp;C[0]</tt> element is a two byte access. If size information wasn't
172available in the query, even the first case would have to conservatively assume
173that the accesses alias.</p>
174
175</div>
176
177<!-- ======================================================================= -->
178<div class="doc_subsection">
179 <a name="alias">The <tt>alias</tt> method</a>
180</div>
181
182<div class="doc_text">
183The <tt>alias</tt> method is the primary interface used to determine whether or
184not two memory objects alias each other. It takes two memory objects as input
185and returns MustAlias, MayAlias, or NoAlias as appropriate.
186</div>
187
188<!-- _______________________________________________________________________ -->
189<div class="doc_subsubsection">
190 <a name="MustMayNo">Must, May, and No Alias Responses</a>
191</div>
192
193<div class="doc_text">
194
195<p>An Alias Analysis implementation can return one of three responses:
196MustAlias, MayAlias, and NoAlias. The No and May alias results are obvious: if
197the two pointers can never equal each other, return NoAlias, if they might,
198return MayAlias.</p>
199
200<p>The MustAlias response is trickier though. In LLVM, the Must Alias response
201may only be returned if the two memory objects are guaranteed to always start at
202exactly the same location. If two memory objects overlap, but do not start at
203the same location, return MayAlias.</p>
204
205</div>
206
207<!-- ======================================================================= -->
208<div class="doc_subsection">
209 <a name="ModRefInfo">The <tt>getModRefInfo</tt> methods</a>
210</div>
211
212<div class="doc_text">
213
214<p>The <tt>getModRefInfo</tt> methods return information about whether the
215execution of an instruction can read or modify a memory location. Mod/Ref
216information is always conservative: if an instruction <b>might</b> read or write
217a location, ModRef is returned.</p>
218
219<p>The <tt>AliasAnalysis</tt> class also provides a <tt>getModRefInfo</tt>
220method for testing dependencies between function calls. This method takes two
221call sites (CS1 &amp; CS2), returns NoModRef if the two calls refer to disjoint
222memory locations, Ref if CS1 reads memory written by CS2, Mod if CS1 writes to
223memory read or written by CS2, or ModRef if CS1 might read or write memory
224accessed by CS2. Note that this relation is not commutative. Clients that use
225this method should be predicated on the <tt>hasNoModRefInfoForCalls()</tt>
226method, which indicates whether or not an analysis can provide mod/ref
227information for function call pairs (most can not). If this predicate is false,
228the client shouldn't waste analysis time querying the <tt>getModRefInfo</tt>
229method many times.</p>
230
231</div>
232
233
234<!-- ======================================================================= -->
235<div class="doc_subsection">
236 <a name="OtherItfs">Other useful <tt>AliasAnalysis</tt> methods</a>
237</div>
238
239<div class="doc_text">
240
241<p>
242Several other tidbits of information are often collected by various alias
243analysis implementations and can be put to good use by various clients.
244</p>
245
246</div>
247
248<!-- _______________________________________________________________________ -->
249<div class="doc_subsubsection">
250 The <tt>getMustAliases</tt> method
251</div>
252
253<div class="doc_text">
254
255<p>The <tt>getMustAliases</tt> method returns all values that are known to
256always must alias a pointer. This information can be provided in some cases for
257important objects like the null pointer and global values. Knowing that a
258pointer always points to a particular function allows indirect calls to be
259turned into direct calls, for example.</p>
260
261</div>
262
263<!-- _______________________________________________________________________ -->
264<div class="doc_subsubsection">
265 The <tt>pointsToConstantMemory</tt> method
266</div>
267
268<div class="doc_text">
269
270<p>The <tt>pointsToConstantMemory</tt> method returns true if and only if the
271analysis can prove that the pointer only points to unchanging memory locations
272(functions, constant global variables, and the null pointer). This information
273can be used to refine mod/ref information: it is impossible for an unchanging
274memory location to be modified.</p>
275
276</div>
277
278<!-- _______________________________________________________________________ -->
279<div class="doc_subsubsection">
280 <a name="simplemodref">The <tt>doesNotAccessMemory</tt> and
281 <tt>onlyReadsMemory</tt> methods</a>
282</div>
283
284<div class="doc_text">
285
286<p>These methods are used to provide very simple mod/ref information for
287function calls. The <tt>doesNotAccessMemory</tt> method returns true for a
288function if the analysis can prove that the function never reads or writes to
289memory, or if the function only reads from constant memory. Functions with this
290property are side-effect free and only depend on their input arguments, allowing
291them to be eliminated if they form common subexpressions or be hoisted out of
292loops. Many common functions behave this way (e.g., <tt>sin</tt> and
293<tt>cos</tt>) but many others do not (e.g., <tt>acos</tt>, which modifies the
294<tt>errno</tt> variable).</p>
295
296<p>The <tt>onlyReadsMemory</tt> method returns true for a function if analysis
297can prove that (at most) the function only reads from non-volatile memory.
298Functions with this property are side-effect free, only depending on their input
299arguments and the state of memory when they are called. This property allows
300calls to these functions to be eliminated and moved around, as long as there is
301no store instruction that changes the contents of memory. Note that all
302functions that satisfy the <tt>doesNotAccessMemory</tt> method also satisfies
303<tt>onlyReadsMemory</tt>.</p>
304
305</div>
306
307<!-- *********************************************************************** -->
308<div class="doc_section">
309 <a name="writingnew">Writing a new <tt>AliasAnalysis</tt> Implementation</a>
310</div>
311<!-- *********************************************************************** -->
312
313<div class="doc_text">
314
315<p>Writing a new alias analysis implementation for LLVM is quite
316straight-forward. There are already several implementations that you can use
317for examples, and the following information should help fill in any details.
318For a examples, take a look at the <a href="#impls">various alias analysis
319implementations</a> included with LLVM.</p>
320
321</div>
322
323<!-- ======================================================================= -->
324<div class="doc_subsection">
325 <a name="passsubclasses">Different Pass styles</a>
326</div>
327
328<div class="doc_text">
329
330<p>The first step to determining what type of <a
331href="WritingAnLLVMPass.html">LLVM pass</a> you need to use for your Alias
332Analysis. As is the case with most other analyses and transformations, the
333answer should be fairly obvious from what type of problem you are trying to
334solve:</p>
335
336<ol>
337 <li>If you require interprocedural analysis, it should be a
338 <tt>Pass</tt>.</li>
339 <li>If you are a function-local analysis, subclass <tt>FunctionPass</tt>.</li>
340 <li>If you don't need to look at the program at all, subclass
341 <tt>ImmutablePass</tt>.</li>
342</ol>
343
344<p>In addition to the pass that you subclass, you should also inherit from the
345<tt>AliasAnalysis</tt> interface, of course, and use the
346<tt>RegisterAnalysisGroup</tt> template to register as an implementation of
347<tt>AliasAnalysis</tt>.</p>
348
349</div>
350
351<!-- ======================================================================= -->
352<div class="doc_subsection">
353 <a name="requiredcalls">Required initialization calls</a>
354</div>
355
356<div class="doc_text">
357
358<p>Your subclass of <tt>AliasAnalysis</tt> is required to invoke two methods on
359the <tt>AliasAnalysis</tt> base class: <tt>getAnalysisUsage</tt> and
360<tt>InitializeAliasAnalysis</tt>. In particular, your implementation of
361<tt>getAnalysisUsage</tt> should explicitly call into the
362<tt>AliasAnalysis::getAnalysisUsage</tt> method in addition to doing any
363declaring any pass dependencies your pass has. Thus you should have something
364like this:</p>
365
366<div class="doc_code">
367<pre>
368void getAnalysisUsage(AnalysisUsage &amp;AU) const {
369 AliasAnalysis::getAnalysisUsage(AU);
370 <i>// declare your dependencies here.</i>
371}
372</pre>
373</div>
374
375<p>Additionally, your must invoke the <tt>InitializeAliasAnalysis</tt> method
376from your analysis run method (<tt>run</tt> for a <tt>Pass</tt>,
377<tt>runOnFunction</tt> for a <tt>FunctionPass</tt>, or <tt>InitializePass</tt>
378for an <tt>ImmutablePass</tt>). For example (as part of a <tt>Pass</tt>):</p>
379
380<div class="doc_code">
381<pre>
382bool run(Module &amp;M) {
383 InitializeAliasAnalysis(this);
384 <i>// Perform analysis here...</i>
385 return false;
386}
387</pre>
388</div>
389
390</div>
391
392<!-- ======================================================================= -->
393<div class="doc_subsection">
394 <a name="interfaces">Interfaces which may be specified</a>
395</div>
396
397<div class="doc_text">
398
399<p>All of the <a
400href="/doxygen/classllvm_1_1AliasAnalysis.html"><tt>AliasAnalysis</tt></a>
401virtual methods default to providing <a href="#chaining">chaining</a> to another
402alias analysis implementation, which ends up returning conservatively correct
403information (returning "May" Alias and "Mod/Ref" for alias and mod/ref queries
404respectively). Depending on the capabilities of the analysis you are
405implementing, you just override the interfaces you can improve.</p>
406
407</div>
408
409
410
411<!-- ======================================================================= -->
412<div class="doc_subsection">
413 <a name="chaining"><tt>AliasAnalysis</tt> chaining behavior</a>
414</div>
415
416<div class="doc_text">
417
418<p>With only two special exceptions (the <tt><a
419href="#basic-aa">basicaa</a></tt> and <a href="#no-aa"><tt>no-aa</tt></a>
420passes) every alias analysis pass chains to another alias analysis
421implementation (for example, the user can specify "<tt>-basicaa -ds-aa
422-anders-aa -licm</tt>" to get the maximum benefit from the three alias
423analyses). The alias analysis class automatically takes care of most of this
424for methods that you don't override. For methods that you do override, in code
425paths that return a conservative MayAlias or Mod/Ref result, simply return
426whatever the superclass computes. For example:</p>
427
428<div class="doc_code">
429<pre>
430AliasAnalysis::AliasResult alias(const Value *V1, unsigned V1Size,
431 const Value *V2, unsigned V2Size) {
432 if (...)
433 return NoAlias;
434 ...
435
436 <i>// Couldn't determine a must or no-alias result.</i>
437 return AliasAnalysis::alias(V1, V1Size, V2, V2Size);
438}
439</pre>
440</div>
441
442<p>In addition to analysis queries, you must make sure to unconditionally pass
443LLVM <a href="#updating">update notification</a> methods to the superclass as
444well if you override them, which allows all alias analyses in a change to be
445updated.</p>
446
447</div>
448
449
450<!-- ======================================================================= -->
451<div class="doc_subsection">
452 <a name="updating">Updating analysis results for transformations</a>
453</div>
454
455<div class="doc_text">
456<p>
457Alias analysis information is initially computed for a static snapshot of the
458program, but clients will use this information to make transformations to the
459code. All but the most trivial forms of alias analysis will need to have their
460analysis results updated to reflect the changes made by these transformations.
461</p>
462
463<p>
464The <tt>AliasAnalysis</tt> interface exposes two methods which are used to
465communicate program changes from the clients to the analysis implementations.
466Various alias analysis implementations should use these methods to ensure that
467their internal data structures are kept up-to-date as the program changes (for
468example, when an instruction is deleted), and clients of alias analysis must be
469sure to call these interfaces appropriately.
470</p>
471</div>
472
473<!-- _______________________________________________________________________ -->
474<div class="doc_subsubsection">The <tt>deleteValue</tt> method</div>
475
476<div class="doc_text">
477The <tt>deleteValue</tt> method is called by transformations when they remove an
478instruction or any other value from the program (including values that do not
479use pointers). Typically alias analyses keep data structures that have entries
480for each value in the program. When this method is called, they should remove
481any entries for the specified value, if they exist.
482</div>
483
484<!-- _______________________________________________________________________ -->
485<div class="doc_subsubsection">The <tt>copyValue</tt> method</div>
486
487<div class="doc_text">
488The <tt>copyValue</tt> method is used when a new value is introduced into the
489program. There is no way to introduce a value into the program that did not
490exist before (this doesn't make sense for a safe compiler transformation), so
491this is the only way to introduce a new value. This method indicates that the
492new value has exactly the same properties as the value being copied.
493</div>
494
495<!-- _______________________________________________________________________ -->
496<div class="doc_subsubsection">The <tt>replaceWithNewValue</tt> method</div>
497
498<div class="doc_text">
499This method is a simple helper method that is provided to make clients easier to
500use. It is implemented by copying the old analysis information to the new
501value, then deleting the old value. This method cannot be overridden by alias
502analysis implementations.
503</div>
504
505<!-- ======================================================================= -->
506<div class="doc_subsection">
507 <a name="implefficiency">Efficiency Issues</a>
508</div>
509
510<div class="doc_text">
511
512<p>From the LLVM perspective, the only thing you need to do to provide an
513efficient alias analysis is to make sure that alias analysis <b>queries</b> are
514serviced quickly. The actual calculation of the alias analysis results (the
515"run" method) is only performed once, but many (perhaps duplicate) queries may
516be performed. Because of this, try to move as much computation to the run
517method as possible (within reason).</p>
518
519</div>
520
521<!-- *********************************************************************** -->
522<div class="doc_section">
523 <a name="using">Using alias analysis results</a>
524</div>
525<!-- *********************************************************************** -->
526
527<div class="doc_text">
528
529<p>There are several different ways to use alias analysis results. In order of
530preference, these are...</p>
531
532</div>
533
534<!-- ======================================================================= -->
535<div class="doc_subsection">
536 <a name="loadvn">Using the <tt>-load-vn</tt> Pass</a>
537</div>
538
539<div class="doc_text">
540
541<p>The <tt>load-vn</tt> pass uses alias analysis to provide value numbering
542information for <tt>load</tt> instructions and pointer values. If your analysis
543or transformation can be modeled in a form that uses value numbering
544information, you don't have to do anything special to handle load instructions:
545just use the <tt>load-vn</tt> pass, which uses alias analysis.</p>
546
547</div>
548
549<!-- ======================================================================= -->
550<div class="doc_subsection">
551 <a name="ast">Using the <tt>AliasSetTracker</tt> class</a>
552</div>
553
554<div class="doc_text">
555
556<p>Many transformations need information about alias <b>sets</b> that are active
557in some scope, rather than information about pairwise aliasing. The <tt><a
558href="/doxygen/classllvm_1_1AliasSetTracker.html">AliasSetTracker</a></tt> class
559is used to efficiently build these Alias Sets from the pairwise alias analysis
560information provided by the <tt>AliasAnalysis</tt> interface.</p>
561
562<p>First you initialize the AliasSetTracker by using the "<tt>add</tt>" methods
563to add information about various potentially aliasing instructions in the scope
564you are interested in. Once all of the alias sets are completed, your pass
565should simply iterate through the constructed alias sets, using the
566<tt>AliasSetTracker</tt> <tt>begin()</tt>/<tt>end()</tt> methods.</p>
567
568<p>The <tt>AliasSet</tt>s formed by the <tt>AliasSetTracker</tt> are guaranteed
569to be disjoint, calculate mod/ref information and volatility for the set, and
570keep track of whether or not all of the pointers in the set are Must aliases.
571The AliasSetTracker also makes sure that sets are properly folded due to call
572instructions, and can provide a list of pointers in each set.</p>
573
574<p>As an example user of this, the <a href="/doxygen/structLICM.html">Loop
575Invariant Code Motion</a> pass uses <tt>AliasSetTracker</tt>s to calculate alias
576sets for each loop nest. If an <tt>AliasSet</tt> in a loop is not modified,
577then all load instructions from that set may be hoisted out of the loop. If any
578alias sets are stored to <b>and</b> are must alias sets, then the stores may be
579sunk to outside of the loop, promoting the memory location to a register for the
580duration of the loop nest. Both of these transformations only apply if the
581pointer argument is loop-invariant.</p>
582
583</div>
584
585<!-- _______________________________________________________________________ -->
586<div class="doc_subsubsection">
587 The AliasSetTracker implementation
588</div>
589
590<div class="doc_text">
591
592<p>The AliasSetTracker class is implemented to be as efficient as possible. It
593uses the union-find algorithm to efficiently merge AliasSets when a pointer is
594inserted into the AliasSetTracker that aliases multiple sets. The primary data
595structure is a hash table mapping pointers to the AliasSet they are in.</p>
596
597<p>The AliasSetTracker class must maintain a list of all of the LLVM Value*'s
598that are in each AliasSet. Since the hash table already has entries for each
599LLVM Value* of interest, the AliasesSets thread the linked list through these
600hash-table nodes to avoid having to allocate memory unnecessarily, and to make
601merging alias sets extremely efficient (the linked list merge is constant time).
602</p>
603
604<p>You shouldn't need to understand these details if you are just a client of
605the AliasSetTracker, but if you look at the code, hopefully this brief
606description will help make sense of why things are designed the way they
607are.</p>
608
609</div>
610
611<!-- ======================================================================= -->
612<div class="doc_subsection">
613 <a name="direct">Using the <tt>AliasAnalysis</tt> interface directly</a>
614</div>
615
616<div class="doc_text">
617
618<p>If neither of these utility class are what your pass needs, you should use
619the interfaces exposed by the <tt>AliasAnalysis</tt> class directly. Try to use
620the higher-level methods when possible (e.g., use mod/ref information instead of
621the <a href="#alias"><tt>alias</tt></a> method directly if possible) to get the
622best precision and efficiency.</p>
623
624</div>
625
626<!-- *********************************************************************** -->
627<div class="doc_section">
628 <a name="exist">Existing alias analysis implementations and clients</a>
629</div>
630<!-- *********************************************************************** -->
631
632<div class="doc_text">
633
634<p>If you're going to be working with the LLVM alias analysis infrastructure,
635you should know what clients and implementations of alias analysis are
636available. In particular, if you are implementing an alias analysis, you should
637be aware of the <a href="#aliasanalysis-debug">the clients</a> that are useful
638for monitoring and evaluating different implementations.</p>
639
640</div>
641
642<!-- ======================================================================= -->
643<div class="doc_subsection">
644 <a name="impls">Available <tt>AliasAnalysis</tt> implementations</a>
645</div>
646
647<div class="doc_text">
648
649<p>This section lists the various implementations of the <tt>AliasAnalysis</tt>
650interface. With the exception of the <a href="#no-aa"><tt>-no-aa</tt></a> and
651<a href="#basic-aa"><tt>-basicaa</tt></a> implementations, all of these <a
652href="#chaining">chain</a> to other alias analysis implementations.</p>
653
654</div>
655
656<!-- _______________________________________________________________________ -->
657<div class="doc_subsubsection">
658 <a name="no-aa">The <tt>-no-aa</tt> pass</a>
659</div>
660
661<div class="doc_text">
662
663<p>The <tt>-no-aa</tt> pass is just like what it sounds: an alias analysis that
664never returns any useful information. This pass can be useful if you think that
665alias analysis is doing something wrong and are trying to narrow down a
666problem.</p>
667
668</div>
669
670<!-- _______________________________________________________________________ -->
671<div class="doc_subsubsection">
672 <a name="basic-aa">The <tt>-basicaa</tt> pass</a>
673</div>
674
675<div class="doc_text">
676
677<p>The <tt>-basicaa</tt> pass is the default LLVM alias analysis. It is an
678aggressive local analysis that "knows" many important facts:</p>
679
680<ul>
681<li>Distinct globals, stack allocations, and heap allocations can never
682 alias.</li>
683<li>Globals, stack allocations, and heap allocations never alias the null
684 pointer.</li>
685<li>Different fields of a structure do not alias.</li>
686<li>Indexes into arrays with statically differing subscripts cannot alias.</li>
687<li>Many common standard C library functions <a
688 href="#simplemodref">never access memory or only read memory</a>.</li>
689<li>Pointers that obviously point to constant globals
690 "<tt>pointToConstantMemory</tt>".</li>
691<li>Function calls can not modify or references stack allocations if they never
692 escape from the function that allocates them (a common case for automatic
693 arrays).</li>
694</ul>
695
696</div>
697
698<!-- _______________________________________________________________________ -->
699<div class="doc_subsubsection">
700 <a name="globalsmodref">The <tt>-globalsmodref-aa</tt> pass</a>
701</div>
702
703<div class="doc_text">
704
705<p>This pass implements a simple context-sensitive mod/ref and alias analysis
706for internal global variables that don't "have their address taken". If a
707global does not have its address taken, the pass knows that no pointers alias
708the global. This pass also keeps track of functions that it knows never access
709memory or never read memory. This allows certain optimizations (e.g. GCSE) to
710eliminate call instructions entirely.
711</p>
712
713<p>The real power of this pass is that it provides context-sensitive mod/ref
714information for call instructions. This allows the optimizer to know that
715calls to a function do not clobber or read the value of the global, allowing
716loads and stores to be eliminated.</p>
717
718<p>Note that this pass is somewhat limited in its scope (only support
719non-address taken globals), but is very quick analysis.</p>
720</div>
721
722<!-- _______________________________________________________________________ -->
723<div class="doc_subsubsection">
724 <a name="anders-aa">The <tt>-anders-aa</tt> pass</a>
725</div>
726
727<div class="doc_text">
728
729<p>The <tt>-anders-aa</tt> pass implements the well-known "Andersen's algorithm"
730for interprocedural alias analysis. This algorithm is a subset-based,
731flow-insensitive, context-insensitive, and field-insensitive alias analysis that
732is widely believed to be fairly precise. Unfortunately, this algorithm is also
733O(N<sup>3</sup>). The LLVM implementation currently does not implement any of
734the refinements (such as "online cycle elimination" or "offline variable
735substitution") to improve its efficiency, so it can be quite slow in common
736cases.
737</p>
738
739</div>
740
741<!-- _______________________________________________________________________ -->
742<div class="doc_subsubsection">
743 <a name="steens-aa">The <tt>-steens-aa</tt> pass</a>
744</div>
745
746<div class="doc_text">
747
748<p>The <tt>-steens-aa</tt> pass implements a variation on the well-known
749"Steensgaard's algorithm" for interprocedural alias analysis. Steensgaard's
750algorithm is a unification-based, flow-insensitive, context-insensitive, and
751field-insensitive alias analysis that is also very scalable (effectively linear
752time).</p>
753
754<p>The LLVM <tt>-steens-aa</tt> pass implements a "speculatively
755field-<b>sensitive</b>" version of Steensgaard's algorithm using the Data
756Structure Analysis framework. This gives it substantially more precision than
757the standard algorithm while maintaining excellent analysis scalability.</p>
758
759<p>Note that <tt>-steens-aa</tt> is available in the optional "poolalloc"
760module, it is not part of the LLVM core.</p>
761
762</div>
763
764<!-- _______________________________________________________________________ -->
765<div class="doc_subsubsection">
766 <a name="ds-aa">The <tt>-ds-aa</tt> pass</a>
767</div>
768
769<div class="doc_text">
770
771<p>The <tt>-ds-aa</tt> pass implements the full Data Structure Analysis
772algorithm. Data Structure Analysis is a modular unification-based,
773flow-insensitive, context-<b>sensitive</b>, and speculatively
774field-<b>sensitive</b> alias analysis that is also quite scalable, usually at
775O(n*log(n)).</p>
776
777<p>This algorithm is capable of responding to a full variety of alias analysis
778queries, and can provide context-sensitive mod/ref information as well. The
779only major facility not implemented so far is support for must-alias
780information.</p>
781
782<p>Note that <tt>-ds-aa</tt> is available in the optional "poolalloc"
783module, it is not part of the LLVM core.</p>
784
785</div>
786
787
788<!-- ======================================================================= -->
789<div class="doc_subsection">
790 <a name="aliasanalysis-xforms">Alias analysis driven transformations</a>
791</div>
792
793<div class="doc_text">
794LLVM includes several alias-analysis driven transformations which can be used
795with any of the implementations above.
796</div>
797
798<!-- _______________________________________________________________________ -->
799<div class="doc_subsubsection">
800 <a name="adce">The <tt>-adce</tt> pass</a>
801</div>
802
803<div class="doc_text">
804
805<p>The <tt>-adce</tt> pass, which implements Aggressive Dead Code Elimination
806uses the <tt>AliasAnalysis</tt> interface to delete calls to functions that do
807not have side-effects and are not used.</p>
808
809</div>
810
811
812<!-- _______________________________________________________________________ -->
813<div class="doc_subsubsection">
814 <a name="licm">The <tt>-licm</tt> pass</a>
815</div>
816
817<div class="doc_text">
818
819<p>The <tt>-licm</tt> pass implements various Loop Invariant Code Motion related
820transformations. It uses the <tt>AliasAnalysis</tt> interface for several
821different transformations:</p>
822
823<ul>
824<li>It uses mod/ref information to hoist or sink load instructions out of loops
825if there are no instructions in the loop that modifies the memory loaded.</li>
826
827<li>It uses mod/ref information to hoist function calls out of loops that do not
828write to memory and are loop-invariant.</li>
829
830<li>If uses alias information to promote memory objects that are loaded and
831stored to in loops to live in a register instead. It can do this if there are
832no may aliases to the loaded/stored memory location.</li>
833</ul>
834
835</div>
836
837<!-- _______________________________________________________________________ -->
838<div class="doc_subsubsection">
839 <a name="argpromotion">The <tt>-argpromotion</tt> pass</a>
840</div>
841
842<div class="doc_text">
843<p>
844The <tt>-argpromotion</tt> pass promotes by-reference arguments to be passed in
845by-value instead. In particular, if pointer arguments are only loaded from it
846passes in the value loaded instead of the address to the function. This pass
847uses alias information to make sure that the value loaded from the argument
848pointer is not modified between the entry of the function and any load of the
849pointer.</p>
850</div>
851
852<!-- _______________________________________________________________________ -->
853<div class="doc_subsubsection">
854 <a name="gcseloadvn">The <tt>-load-vn</tt> &amp; <tt>-gcse</tt> passes</a>
855</div>
856
857<div class="doc_text">
858
859<p>The <tt>-load-vn</tt> pass uses alias analysis to "<a href="#loadvn">value
860number</a>" loads and pointers values, which is used by the GCSE pass to
861eliminate instructions. The <tt>-load-vn</tt> pass relies on alias information
862and must-alias information. This combination of passes can make the following
863transformations:</p>
864
865<ul>
866<li>Redundant load instructions are eliminated.</li>
867<li>Load instructions that follow a store to the same location are replaced with
868the stored value ("store forwarding").</li>
869<li>Pointers values (e.g. formal arguments) that must-alias simpler expressions
870(e.g. global variables or the null pointer) are replaced. Note that this
871implements transformations like "virtual method resolution", turning indirect
872calls into direct calls.</li>
873</ul>
874
875</div>
876
877<!-- ======================================================================= -->
878<div class="doc_subsection">
879 <a name="aliasanalysis-debug">Clients for debugging and evaluation of
880 implementations</a>
881</div>
882
883<div class="doc_text">
884
885<p>These passes are useful for evaluating the various alias analysis
886implementations. You can use them with commands like '<tt>opt -anders-aa -ds-aa
887-aa-eval foo.bc -disable-output -stats</tt>'.</p>
888
889</div>
890
891<!-- _______________________________________________________________________ -->
892<div class="doc_subsubsection">
893 <a name="print-alias-sets">The <tt>-print-alias-sets</tt> pass</a>
894</div>
895
896<div class="doc_text">
897
898<p>The <tt>-print-alias-sets</tt> pass is exposed as part of the
899<tt>opt</tt> tool to print out the Alias Sets formed by the <a
900href="#ast"><tt>AliasSetTracker</tt></a> class. This is useful if you're using
901the <tt>AliasSetTracker</tt> class. To use it, use something like:</p>
902
903<div class="doc_code">
904<pre>
905% opt -ds-aa -print-alias-sets -disable-output
906</pre>
907</div>
908
909</div>
910
911
912<!-- _______________________________________________________________________ -->
913<div class="doc_subsubsection">
914 <a name="count-aa">The <tt>-count-aa</tt> pass</a>
915</div>
916
917<div class="doc_text">
918
919<p>The <tt>-count-aa</tt> pass is useful to see how many queries a particular
920pass is making and what responses are returned by the alias analysis. As an
921example,</p>
922
923<div class="doc_code">
924<pre>
925% opt -basicaa -count-aa -ds-aa -count-aa -licm
926</pre>
927</div>
928
929<p>will print out how many queries (and what responses are returned) by the
930<tt>-licm</tt> pass (of the <tt>-ds-aa</tt> pass) and how many queries are made
931of the <tt>-basicaa</tt> pass by the <tt>-ds-aa</tt> pass. This can be useful
932when debugging a transformation or an alias analysis implementation.</p>
933
934</div>
935
936<!-- _______________________________________________________________________ -->
937<div class="doc_subsubsection">
938 <a name="aa-eval">The <tt>-aa-eval</tt> pass</a>
939</div>
940
941<div class="doc_text">
942
943<p>The <tt>-aa-eval</tt> pass simply iterates through all pairs of pointers in a
944function and asks an alias analysis whether or not the pointers alias. This
945gives an indication of the precision of the alias analysis. Statistics are
946printed indicating the percent of no/may/must aliases found (a more precise
947algorithm will have a lower number of may aliases).</p>
948
949</div>
950
951<!-- *********************************************************************** -->
Owen Anderson05e080f2007-10-02 00:43:25 +0000[diff] [blame^]952<div class="doc_section">
953 <a name="memdep">Memory Dependence Analysis</a>
954</div>
955<!-- *********************************************************************** -->
956
957<div class="doc_text">
958
959<p>If you're just looking to be a client of alias analysis information, consider
960using the Memory Dependence Analysis interface instead. MemDep is a lazy,
961caching layer on top of alias analysis that is able to answer the question of
962what preceding memory operations a given instruction depends on, either at an
963intra- or inter-block level. Because of its laziness and caching
964policy, using MemDep can be a significant performance win over accessing alias
965analysis directly.</p>
966
967</div>
968
969<!-- *********************************************************************** -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]970
971<hr>
972<address>
973 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
974 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
975 <a href="http://validator.w3.org/check/referer"><img
976 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
977
978 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
979 <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
980 Last modified: $Date$
981</address>
982
983</body>
984</html>