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 <h1> 11 LLVM Alias Analysis Infrastructure 12 </h1> 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 <li><a href="#limitations">Limitations</a></li> 35 </ul> 36 </li> 37 38 <li><a href="#using">Using alias analysis results</a> 39 <ul> 40 <li><a href="#memdep">Using the <tt>MemoryDependenceAnalysis</tt> Pass</a></li> 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> 54 <li><a href="#memdep">Memory Dependence Analysis</a></li> 55 </ol> 56 57 <div class="doc_author"> 58 <p>Written by <a href="mailto:sabre (a] nondot.org">Chris Lattner</a></p> 59 </div> 60 61 <!-- *********************************************************************** --> 62 <h2> 63 <a name="introduction">Introduction</a> 64 </h2> 65 <!-- *********************************************************************** --> 66 67 <div> 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 (a] nondot.org">let me 94 know</a>.</p> 95 96 </div> 97 98 <!-- *********************************************************************** --> 99 <h2> 100 <a name="overview"><tt>AliasAnalysis</tt> Class Overview</a> 101 </h2> 102 <!-- *********************************************************************** --> 103 104 <div> 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 <p>All <tt>AliasAnalysis</tt> interfaces require that in queries involving 121 multiple values, values which are not 122 <a href="LangRef.html#constants">constants</a> are all defined within the 123 same function.</p> 124 125 <!-- ======================================================================= --> 126 <h3> 127 <a name="pointers">Representation of Pointers</a> 128 </h3> 129 130 <div> 131 132 <p>Most importantly, the <tt>AliasAnalysis</tt> class provides several methods 133 which are used to query whether or not two memory objects alias, whether 134 function calls can modify or read a memory object, etc. For all of these 135 queries, memory objects are represented as a pair of their starting address (a 136 symbolic LLVM <tt>Value*</tt>) and a static size.</p> 137 138 <p>Representing memory objects as a starting address and a size is critically 139 important for correct Alias Analyses. For example, consider this (silly, but 140 possible) C code:</p> 141 142 <div class="doc_code"> 143 <pre> 144 int i; 145 char C[2]; 146 char A[10]; 147 /* ... */ 148 for (i = 0; i != 10; ++i) { 149 C[0] = A[i]; /* One byte store */ 150 C[1] = A[9-i]; /* One byte store */ 151 } 152 </pre> 153 </div> 154 155 <p>In this case, the <tt>basicaa</tt> pass will disambiguate the stores to 156 <tt>C[0]</tt> and <tt>C[1]</tt> because they are accesses to two distinct 157 locations one byte apart, and the accesses are each one byte. In this case, the 158 LICM pass can use store motion to remove the stores from the loop. In 159 constrast, the following code:</p> 160 161 <div class="doc_code"> 162 <pre> 163 int i; 164 char C[2]; 165 char A[10]; 166 /* ... */ 167 for (i = 0; i != 10; ++i) { 168 ((short*)C)[0] = A[i]; /* Two byte store! */ 169 C[1] = A[9-i]; /* One byte store */ 170 } 171 </pre> 172 </div> 173 174 <p>In this case, the two stores to C do alias each other, because the access to 175 the <tt>&C[0]</tt> element is a two byte access. If size information wasn't 176 available in the query, even the first case would have to conservatively assume 177 that the accesses alias.</p> 178 179 </div> 180 181 <!-- ======================================================================= --> 182 <h3> 183 <a name="alias">The <tt>alias</tt> method</a> 184 </h3> 185 186 <div> 187 <p>The <tt>alias</tt> method is the primary interface used to determine whether 188 or not two memory objects alias each other. It takes two memory objects as 189 input and returns MustAlias, PartialAlias, MayAlias, or NoAlias as 190 appropriate.</p> 191 192 <p>Like all <tt>AliasAnalysis</tt> interfaces, the <tt>alias</tt> method requires 193 that either the two pointer values be defined within the same function, or at 194 least one of the values is a <a href="LangRef.html#constants">constant</a>.</p> 195 196 <!-- _______________________________________________________________________ --> 197 <h4> 198 <a name="MustMayNo">Must, May, and No Alias Responses</a> 199 </h4> 200 201 <div> 202 <p>The NoAlias response may be used when there is never an immediate dependence 203 between any memory reference <i>based</i> on one pointer and any memory 204 reference <i>based</i> the other. The most obvious example is when the two 205 pointers point to non-overlapping memory ranges. Another is when the two 206 pointers are only ever used for reading memory. Another is when the memory is 207 freed and reallocated between accesses through one pointer and accesses through 208 the other -- in this case, there is a dependence, but it's mediated by the free 209 and reallocation.</p> 210 211 <p>As an exception to this is with the 212 <a href="LangRef.html#noalias"><tt>noalias</tt></a> keyword; the "irrelevant" 213 dependencies are ignored.</p> 214 215 <p>The MayAlias response is used whenever the two pointers might refer to the 216 same object.</p> 217 218 <p>The PartialAlias response is used when the two memory objects are known 219 to be overlapping in some way, but do not start at the same address.</p> 220 221 <p>The MustAlias response may only be returned if the two memory objects are 222 guaranteed to always start at exactly the same location. A MustAlias response 223 implies that the pointers compare equal.</p> 224 225 </div> 226 227 </div> 228 229 <!-- ======================================================================= --> 230 <h3> 231 <a name="ModRefInfo">The <tt>getModRefInfo</tt> methods</a> 232 </h3> 233 234 <div> 235 236 <p>The <tt>getModRefInfo</tt> methods return information about whether the 237 execution of an instruction can read or modify a memory location. Mod/Ref 238 information is always conservative: if an instruction <b>might</b> read or write 239 a location, ModRef is returned.</p> 240 241 <p>The <tt>AliasAnalysis</tt> class also provides a <tt>getModRefInfo</tt> 242 method for testing dependencies between function calls. This method takes two 243 call sites (CS1 & CS2), returns NoModRef if neither call writes to memory 244 read or written by the other, Ref if CS1 reads memory written by CS2, Mod if CS1 245 writes to memory read or written by CS2, or ModRef if CS1 might read or write 246 memory written to by CS2. Note that this relation is not commutative.</p> 247 248 </div> 249 250 251 <!-- ======================================================================= --> 252 <h3> 253 <a name="OtherItfs">Other useful <tt>AliasAnalysis</tt> methods</a> 254 </h3> 255 256 <div> 257 258 <p> 259 Several other tidbits of information are often collected by various alias 260 analysis implementations and can be put to good use by various clients. 261 </p> 262 263 <!-- _______________________________________________________________________ --> 264 <h4> 265 The <tt>pointsToConstantMemory</tt> method 266 </h4> 267 268 <div> 269 270 <p>The <tt>pointsToConstantMemory</tt> method returns true if and only if the 271 analysis can prove that the pointer only points to unchanging memory locations 272 (functions, constant global variables, and the null pointer). This information 273 can be used to refine mod/ref information: it is impossible for an unchanging 274 memory location to be modified.</p> 275 276 </div> 277 278 <!-- _______________________________________________________________________ --> 279 <h4> 280 <a name="simplemodref">The <tt>doesNotAccessMemory</tt> and 281 <tt>onlyReadsMemory</tt> methods</a> 282 </h4> 283 284 <div> 285 286 <p>These methods are used to provide very simple mod/ref information for 287 function calls. The <tt>doesNotAccessMemory</tt> method returns true for a 288 function if the analysis can prove that the function never reads or writes to 289 memory, or if the function only reads from constant memory. Functions with this 290 property are side-effect free and only depend on their input arguments, allowing 291 them to be eliminated if they form common subexpressions or be hoisted out of 292 loops. 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 297 can prove that (at most) the function only reads from non-volatile memory. 298 Functions with this property are side-effect free, only depending on their input 299 arguments and the state of memory when they are called. This property allows 300 calls to these functions to be eliminated and moved around, as long as there is 301 no store instruction that changes the contents of memory. Note that all 302 functions that satisfy the <tt>doesNotAccessMemory</tt> method also satisfies 303 <tt>onlyReadsMemory</tt>.</p> 304 305 </div> 306 307 </div> 308 309 </div> 310 311 <!-- *********************************************************************** --> 312 <h2> 313 <a name="writingnew">Writing a new <tt>AliasAnalysis</tt> Implementation</a> 314 </h2> 315 <!-- *********************************************************************** --> 316 317 <div> 318 319 <p>Writing a new alias analysis implementation for LLVM is quite 320 straight-forward. There are already several implementations that you can use 321 for examples, and the following information should help fill in any details. 322 For a examples, take a look at the <a href="#impls">various alias analysis 323 implementations</a> included with LLVM.</p> 324 325 <!-- ======================================================================= --> 326 <h3> 327 <a name="passsubclasses">Different Pass styles</a> 328 </h3> 329 330 <div> 331 332 <p>The first step to determining what type of <a 333 href="WritingAnLLVMPass.html">LLVM pass</a> you need to use for your Alias 334 Analysis. As is the case with most other analyses and transformations, the 335 answer should be fairly obvious from what type of problem you are trying to 336 solve:</p> 337 338 <ol> 339 <li>If you require interprocedural analysis, it should be a 340 <tt>Pass</tt>.</li> 341 <li>If you are a function-local analysis, subclass <tt>FunctionPass</tt>.</li> 342 <li>If you don't need to look at the program at all, subclass 343 <tt>ImmutablePass</tt>.</li> 344 </ol> 345 346 <p>In addition to the pass that you subclass, you should also inherit from the 347 <tt>AliasAnalysis</tt> interface, of course, and use the 348 <tt>RegisterAnalysisGroup</tt> template to register as an implementation of 349 <tt>AliasAnalysis</tt>.</p> 350 351 </div> 352 353 <!-- ======================================================================= --> 354 <h3> 355 <a name="requiredcalls">Required initialization calls</a> 356 </h3> 357 358 <div> 359 360 <p>Your subclass of <tt>AliasAnalysis</tt> is required to invoke two methods on 361 the <tt>AliasAnalysis</tt> base class: <tt>getAnalysisUsage</tt> and 362 <tt>InitializeAliasAnalysis</tt>. In particular, your implementation of 363 <tt>getAnalysisUsage</tt> should explicitly call into the 364 <tt>AliasAnalysis::getAnalysisUsage</tt> method in addition to doing any 365 declaring any pass dependencies your pass has. Thus you should have something 366 like this:</p> 367 368 <div class="doc_code"> 369 <pre> 370 void getAnalysisUsage(AnalysisUsage &AU) const { 371 AliasAnalysis::getAnalysisUsage(AU); 372 <i>// declare your dependencies here.</i> 373 } 374 </pre> 375 </div> 376 377 <p>Additionally, your must invoke the <tt>InitializeAliasAnalysis</tt> method 378 from your analysis run method (<tt>run</tt> for a <tt>Pass</tt>, 379 <tt>runOnFunction</tt> for a <tt>FunctionPass</tt>, or <tt>InitializePass</tt> 380 for an <tt>ImmutablePass</tt>). For example (as part of a <tt>Pass</tt>):</p> 381 382 <div class="doc_code"> 383 <pre> 384 bool run(Module &M) { 385 InitializeAliasAnalysis(this); 386 <i>// Perform analysis here...</i> 387 return false; 388 } 389 </pre> 390 </div> 391 392 </div> 393 394 <!-- ======================================================================= --> 395 <h3> 396 <a name="interfaces">Interfaces which may be specified</a> 397 </h3> 398 399 <div> 400 401 <p>All of the <a 402 href="/doxygen/classllvm_1_1AliasAnalysis.html"><tt>AliasAnalysis</tt></a> 403 virtual methods default to providing <a href="#chaining">chaining</a> to another 404 alias analysis implementation, which ends up returning conservatively correct 405 information (returning "May" Alias and "Mod/Ref" for alias and mod/ref queries 406 respectively). Depending on the capabilities of the analysis you are 407 implementing, you just override the interfaces you can improve.</p> 408 409 </div> 410 411 412 413 <!-- ======================================================================= --> 414 <h3> 415 <a name="chaining"><tt>AliasAnalysis</tt> chaining behavior</a> 416 </h3> 417 418 <div> 419 420 <p>With only two special exceptions (the <tt><a 421 href="#basic-aa">basicaa</a></tt> and <a href="#no-aa"><tt>no-aa</tt></a> 422 passes) every alias analysis pass chains to another alias analysis 423 implementation (for example, the user can specify "<tt>-basicaa -ds-aa 424 -licm</tt>" to get the maximum benefit from both alias 425 analyses). The alias analysis class automatically takes care of most of this 426 for methods that you don't override. For methods that you do override, in code 427 paths that return a conservative MayAlias or Mod/Ref result, simply return 428 whatever the superclass computes. For example:</p> 429 430 <div class="doc_code"> 431 <pre> 432 AliasAnalysis::AliasResult alias(const Value *V1, unsigned V1Size, 433 const Value *V2, unsigned V2Size) { 434 if (...) 435 return NoAlias; 436 ... 437 438 <i>// Couldn't determine a must or no-alias result.</i> 439 return AliasAnalysis::alias(V1, V1Size, V2, V2Size); 440 } 441 </pre> 442 </div> 443 444 <p>In addition to analysis queries, you must make sure to unconditionally pass 445 LLVM <a href="#updating">update notification</a> methods to the superclass as 446 well if you override them, which allows all alias analyses in a change to be 447 updated.</p> 448 449 </div> 450 451 452 <!-- ======================================================================= --> 453 <h3> 454 <a name="updating">Updating analysis results for transformations</a> 455 </h3> 456 457 <div> 458 <p> 459 Alias analysis information is initially computed for a static snapshot of the 460 program, but clients will use this information to make transformations to the 461 code. All but the most trivial forms of alias analysis will need to have their 462 analysis results updated to reflect the changes made by these transformations. 463 </p> 464 465 <p> 466 The <tt>AliasAnalysis</tt> interface exposes four methods which are used to 467 communicate program changes from the clients to the analysis implementations. 468 Various alias analysis implementations should use these methods to ensure that 469 their internal data structures are kept up-to-date as the program changes (for 470 example, when an instruction is deleted), and clients of alias analysis must be 471 sure to call these interfaces appropriately. 472 </p> 473 474 <!-- _______________________________________________________________________ --> 475 <h4>The <tt>deleteValue</tt> method</h4> 476 477 <div> 478 The <tt>deleteValue</tt> method is called by transformations when they remove an 479 instruction or any other value from the program (including values that do not 480 use pointers). Typically alias analyses keep data structures that have entries 481 for each value in the program. When this method is called, they should remove 482 any entries for the specified value, if they exist. 483 </div> 484 485 <!-- _______________________________________________________________________ --> 486 <h4>The <tt>copyValue</tt> method</h4> 487 488 <div> 489 The <tt>copyValue</tt> method is used when a new value is introduced into the 490 program. There is no way to introduce a value into the program that did not 491 exist before (this doesn't make sense for a safe compiler transformation), so 492 this is the only way to introduce a new value. This method indicates that the 493 new value has exactly the same properties as the value being copied. 494 </div> 495 496 <!-- _______________________________________________________________________ --> 497 <h4>The <tt>replaceWithNewValue</tt> method</h4> 498 499 <div> 500 This method is a simple helper method that is provided to make clients easier to 501 use. It is implemented by copying the old analysis information to the new 502 value, then deleting the old value. This method cannot be overridden by alias 503 analysis implementations. 504 </div> 505 506 <!-- _______________________________________________________________________ --> 507 <h4>The <tt>addEscapingUse</tt> method</h4> 508 509 <div> 510 <p>The <tt>addEscapingUse</tt> method is used when the uses of a pointer 511 value have changed in ways that may invalidate precomputed analysis information. 512 Implementations may either use this callback to provide conservative responses 513 for points whose uses have change since analysis time, or may recompute some 514 or all of their internal state to continue providing accurate responses.</p> 515 516 <p>In general, any new use of a pointer value is considered an escaping use, 517 and must be reported through this callback, <em>except</em> for the 518 uses below:</p> 519 520 <ul> 521 <li>A <tt>bitcast</tt> or <tt>getelementptr</tt> of the pointer</li> 522 <li>A <tt>store</tt> through the pointer (but not a <tt>store</tt> 523 <em>of</em> the pointer)</li> 524 <li>A <tt>load</tt> through the pointer</li> 525 </ul> 526 </div> 527 528 </div> 529 530 <!-- ======================================================================= --> 531 <h3> 532 <a name="implefficiency">Efficiency Issues</a> 533 </h3> 534 535 <div> 536 537 <p>From the LLVM perspective, the only thing you need to do to provide an 538 efficient alias analysis is to make sure that alias analysis <b>queries</b> are 539 serviced quickly. The actual calculation of the alias analysis results (the 540 "run" method) is only performed once, but many (perhaps duplicate) queries may 541 be performed. Because of this, try to move as much computation to the run 542 method as possible (within reason).</p> 543 544 </div> 545 546 <!-- ======================================================================= --> 547 <h3> 548 <a name="limitations">Limitations</a> 549 </h3> 550 551 <div> 552 553 <p>The AliasAnalysis infrastructure has several limitations which make 554 writing a new <tt>AliasAnalysis</tt> implementation difficult.</p> 555 556 <p>There is no way to override the default alias analysis. It would 557 be very useful to be able to do something like "opt -my-aa -O2" and 558 have it use -my-aa for all passes which need AliasAnalysis, but there 559 is currently no support for that, short of changing the source code 560 and recompiling. Similarly, there is also no way of setting a chain 561 of analyses as the default.</p> 562 563 <p>There is no way for transform passes to declare that they preserve 564 <tt>AliasAnalysis</tt> implementations. The <tt>AliasAnalysis</tt> 565 interface includes <tt>deleteValue</tt> and <tt>copyValue</tt> methods 566 which are intended to allow a pass to keep an AliasAnalysis consistent, 567 however there's no way for a pass to declare in its 568 <tt>getAnalysisUsage</tt> that it does so. Some passes attempt to use 569 <tt>AU.addPreserved<AliasAnalysis></tt>, however this doesn't 570 actually have any effect.</p> 571 572 <p><tt>AliasAnalysisCounter</tt> (<tt>-count-aa</tt>) and <tt>AliasDebugger</tt> 573 (<tt>-debug-aa</tt>) are implemented as <tt>ModulePass</tt> classes, so if your 574 alias analysis uses <tt>FunctionPass</tt>, it won't be able to use 575 these utilities. If you try to use them, the pass manager will 576 silently route alias analysis queries directly to 577 <tt>BasicAliasAnalysis</tt> instead.</p> 578 579 <p>Similarly, the <tt>opt -p</tt> option introduces <tt>ModulePass</tt> 580 passes between each pass, which prevents the use of <tt>FunctionPass</tt> 581 alias analysis passes.</p> 582 583 <p>The <tt>AliasAnalysis</tt> API does have functions for notifying 584 implementations when values are deleted or copied, however these 585 aren't sufficient. There are many other ways that LLVM IR can be 586 modified which could be relevant to <tt>AliasAnalysis</tt> 587 implementations which can not be expressed.</p> 588 589 <p>The <tt>AliasAnalysisDebugger</tt> utility seems to suggest that 590 <tt>AliasAnalysis</tt> implementations can expect that they will be 591 informed of any relevant <tt>Value</tt> before it appears in an 592 alias query. However, popular clients such as <tt>GVN</tt> don't 593 support this, and are known to trigger errors when run with the 594 <tt>AliasAnalysisDebugger</tt>.</p> 595 596 <p>Due to several of the above limitations, the most obvious use for 597 the <tt>AliasAnalysisCounter</tt> utility, collecting stats on all 598 alias queries in a compilation, doesn't work, even if the 599 <tt>AliasAnalysis</tt> implementations don't use <tt>FunctionPass</tt>. 600 There's no way to set a default, much less a default sequence, 601 and there's no way to preserve it.</p> 602 603 <p>The <tt>AliasSetTracker</tt> class (which is used by <tt>LICM</tt> 604 makes a non-deterministic number of alias queries. This can cause stats 605 collected by <tt>AliasAnalysisCounter</tt> to have fluctuations among 606 identical runs, for example. Another consequence is that debugging 607 techniques involving pausing execution after a predetermined number 608 of queries can be unreliable.</p> 609 610 <p>Many alias queries can be reformulated in terms of other alias 611 queries. When multiple <tt>AliasAnalysis</tt> queries are chained together, 612 it would make sense to start those queries from the beginning of the chain, 613 with care taken to avoid infinite looping, however currently an 614 implementation which wants to do this can only start such queries 615 from itself.</p> 616 617 </div> 618 619 </div> 620 621 <!-- *********************************************************************** --> 622 <h2> 623 <a name="using">Using alias analysis results</a> 624 </h2> 625 <!-- *********************************************************************** --> 626 627 <div> 628 629 <p>There are several different ways to use alias analysis results. In order of 630 preference, these are...</p> 631 632 <!-- ======================================================================= --> 633 <h3> 634 <a name="memdep">Using the <tt>MemoryDependenceAnalysis</tt> Pass</a> 635 </h3> 636 637 <div> 638 639 <p>The <tt>memdep</tt> pass uses alias analysis to provide high-level dependence 640 information about memory-using instructions. This will tell you which store 641 feeds into a load, for example. It uses caching and other techniques to be 642 efficient, and is used by Dead Store Elimination, GVN, and memcpy optimizations. 643 </p> 644 645 </div> 646 647 <!-- ======================================================================= --> 648 <h3> 649 <a name="ast">Using the <tt>AliasSetTracker</tt> class</a> 650 </h3> 651 652 <div> 653 654 <p>Many transformations need information about alias <b>sets</b> that are active 655 in some scope, rather than information about pairwise aliasing. The <tt><a 656 href="/doxygen/classllvm_1_1AliasSetTracker.html">AliasSetTracker</a></tt> class 657 is used to efficiently build these Alias Sets from the pairwise alias analysis 658 information provided by the <tt>AliasAnalysis</tt> interface.</p> 659 660 <p>First you initialize the AliasSetTracker by using the "<tt>add</tt>" methods 661 to add information about various potentially aliasing instructions in the scope 662 you are interested in. Once all of the alias sets are completed, your pass 663 should simply iterate through the constructed alias sets, using the 664 <tt>AliasSetTracker</tt> <tt>begin()</tt>/<tt>end()</tt> methods.</p> 665 666 <p>The <tt>AliasSet</tt>s formed by the <tt>AliasSetTracker</tt> are guaranteed 667 to be disjoint, calculate mod/ref information and volatility for the set, and 668 keep track of whether or not all of the pointers in the set are Must aliases. 669 The AliasSetTracker also makes sure that sets are properly folded due to call 670 instructions, and can provide a list of pointers in each set.</p> 671 672 <p>As an example user of this, the <a href="/doxygen/structLICM.html">Loop 673 Invariant Code Motion</a> pass uses <tt>AliasSetTracker</tt>s to calculate alias 674 sets for each loop nest. If an <tt>AliasSet</tt> in a loop is not modified, 675 then all load instructions from that set may be hoisted out of the loop. If any 676 alias sets are stored to <b>and</b> are must alias sets, then the stores may be 677 sunk to outside of the loop, promoting the memory location to a register for the 678 duration of the loop nest. Both of these transformations only apply if the 679 pointer argument is loop-invariant.</p> 680 681 <!-- _______________________________________________________________________ --> 682 <h4> 683 The AliasSetTracker implementation 684 </h4> 685 686 <div> 687 688 <p>The AliasSetTracker class is implemented to be as efficient as possible. It 689 uses the union-find algorithm to efficiently merge AliasSets when a pointer is 690 inserted into the AliasSetTracker that aliases multiple sets. The primary data 691 structure is a hash table mapping pointers to the AliasSet they are in.</p> 692 693 <p>The AliasSetTracker class must maintain a list of all of the LLVM Value*'s 694 that are in each AliasSet. Since the hash table already has entries for each 695 LLVM Value* of interest, the AliasesSets thread the linked list through these 696 hash-table nodes to avoid having to allocate memory unnecessarily, and to make 697 merging alias sets extremely efficient (the linked list merge is constant time). 698 </p> 699 700 <p>You shouldn't need to understand these details if you are just a client of 701 the AliasSetTracker, but if you look at the code, hopefully this brief 702 description will help make sense of why things are designed the way they 703 are.</p> 704 705 </div> 706 707 </div> 708 709 <!-- ======================================================================= --> 710 <h3> 711 <a name="direct">Using the <tt>AliasAnalysis</tt> interface directly</a> 712 </h3> 713 714 <div> 715 716 <p>If neither of these utility class are what your pass needs, you should use 717 the interfaces exposed by the <tt>AliasAnalysis</tt> class directly. Try to use 718 the higher-level methods when possible (e.g., use mod/ref information instead of 719 the <a href="#alias"><tt>alias</tt></a> method directly if possible) to get the 720 best precision and efficiency.</p> 721 722 </div> 723 724 </div> 725 726 <!-- *********************************************************************** --> 727 <h2> 728 <a name="exist">Existing alias analysis implementations and clients</a> 729 </h2> 730 <!-- *********************************************************************** --> 731 732 <div> 733 734 <p>If you're going to be working with the LLVM alias analysis infrastructure, 735 you should know what clients and implementations of alias analysis are 736 available. In particular, if you are implementing an alias analysis, you should 737 be aware of the <a href="#aliasanalysis-debug">the clients</a> that are useful 738 for monitoring and evaluating different implementations.</p> 739 740 <!-- ======================================================================= --> 741 <h3> 742 <a name="impls">Available <tt>AliasAnalysis</tt> implementations</a> 743 </h3> 744 745 <div> 746 747 <p>This section lists the various implementations of the <tt>AliasAnalysis</tt> 748 interface. With the exception of the <a href="#no-aa"><tt>-no-aa</tt></a> 749 implementation, all of these <a href="#chaining">chain</a> to other alias 750 analysis implementations.</p> 751 752 <!-- _______________________________________________________________________ --> 753 <h4> 754 <a name="no-aa">The <tt>-no-aa</tt> pass</a> 755 </h4> 756 757 <div> 758 759 <p>The <tt>-no-aa</tt> pass is just like what it sounds: an alias analysis that 760 never returns any useful information. This pass can be useful if you think that 761 alias analysis is doing something wrong and are trying to narrow down a 762 problem.</p> 763 764 </div> 765 766 <!-- _______________________________________________________________________ --> 767 <h4> 768 <a name="basic-aa">The <tt>-basicaa</tt> pass</a> 769 </h4> 770 771 <div> 772 773 <p>The <tt>-basicaa</tt> pass is an aggressive local analysis that "knows" 774 many important facts:</p> 775 776 <ul> 777 <li>Distinct globals, stack allocations, and heap allocations can never 778 alias.</li> 779 <li>Globals, stack allocations, and heap allocations never alias the null 780 pointer.</li> 781 <li>Different fields of a structure do not alias.</li> 782 <li>Indexes into arrays with statically differing subscripts cannot alias.</li> 783 <li>Many common standard C library functions <a 784 href="#simplemodref">never access memory or only read memory</a>.</li> 785 <li>Pointers that obviously point to constant globals 786 "<tt>pointToConstantMemory</tt>".</li> 787 <li>Function calls can not modify or references stack allocations if they never 788 escape from the function that allocates them (a common case for automatic 789 arrays).</li> 790 </ul> 791 792 </div> 793 794 <!-- _______________________________________________________________________ --> 795 <h4> 796 <a name="globalsmodref">The <tt>-globalsmodref-aa</tt> pass</a> 797 </h4> 798 799 <div> 800 801 <p>This pass implements a simple context-sensitive mod/ref and alias analysis 802 for internal global variables that don't "have their address taken". If a 803 global does not have its address taken, the pass knows that no pointers alias 804 the global. This pass also keeps track of functions that it knows never access 805 memory or never read memory. This allows certain optimizations (e.g. GVN) to 806 eliminate call instructions entirely. 807 </p> 808 809 <p>The real power of this pass is that it provides context-sensitive mod/ref 810 information for call instructions. This allows the optimizer to know that 811 calls to a function do not clobber or read the value of the global, allowing 812 loads and stores to be eliminated.</p> 813 814 <p>Note that this pass is somewhat limited in its scope (only support 815 non-address taken globals), but is very quick analysis.</p> 816 </div> 817 818 <!-- _______________________________________________________________________ --> 819 <h4> 820 <a name="steens-aa">The <tt>-steens-aa</tt> pass</a> 821 </h4> 822 823 <div> 824 825 <p>The <tt>-steens-aa</tt> pass implements a variation on the well-known 826 "Steensgaard's algorithm" for interprocedural alias analysis. Steensgaard's 827 algorithm is a unification-based, flow-insensitive, context-insensitive, and 828 field-insensitive alias analysis that is also very scalable (effectively linear 829 time).</p> 830 831 <p>The LLVM <tt>-steens-aa</tt> pass implements a "speculatively 832 field-<b>sensitive</b>" version of Steensgaard's algorithm using the Data 833 Structure Analysis framework. This gives it substantially more precision than 834 the standard algorithm while maintaining excellent analysis scalability.</p> 835 836 <p>Note that <tt>-steens-aa</tt> is available in the optional "poolalloc" 837 module, it is not part of the LLVM core.</p> 838 839 </div> 840 841 <!-- _______________________________________________________________________ --> 842 <h4> 843 <a name="ds-aa">The <tt>-ds-aa</tt> pass</a> 844 </h4> 845 846 <div> 847 848 <p>The <tt>-ds-aa</tt> pass implements the full Data Structure Analysis 849 algorithm. Data Structure Analysis is a modular unification-based, 850 flow-insensitive, context-<b>sensitive</b>, and speculatively 851 field-<b>sensitive</b> alias analysis that is also quite scalable, usually at 852 O(n*log(n)).</p> 853 854 <p>This algorithm is capable of responding to a full variety of alias analysis 855 queries, and can provide context-sensitive mod/ref information as well. The 856 only major facility not implemented so far is support for must-alias 857 information.</p> 858 859 <p>Note that <tt>-ds-aa</tt> is available in the optional "poolalloc" 860 module, it is not part of the LLVM core.</p> 861 862 </div> 863 864 <!-- _______________________________________________________________________ --> 865 <h4> 866 <a name="scev-aa">The <tt>-scev-aa</tt> pass</a> 867 </h4> 868 869 <div> 870 871 <p>The <tt>-scev-aa</tt> pass implements AliasAnalysis queries by 872 translating them into ScalarEvolution queries. This gives it a 873 more complete understanding of <tt>getelementptr</tt> instructions 874 and loop induction variables than other alias analyses have.</p> 875 876 </div> 877 878 </div> 879 880 <!-- ======================================================================= --> 881 <h3> 882 <a name="aliasanalysis-xforms">Alias analysis driven transformations</a> 883 </h3> 884 885 <div> 886 LLVM includes several alias-analysis driven transformations which can be used 887 with any of the implementations above. 888 889 <!-- _______________________________________________________________________ --> 890 <h4> 891 <a name="adce">The <tt>-adce</tt> pass</a> 892 </h4> 893 894 <div> 895 896 <p>The <tt>-adce</tt> pass, which implements Aggressive Dead Code Elimination 897 uses the <tt>AliasAnalysis</tt> interface to delete calls to functions that do 898 not have side-effects and are not used.</p> 899 900 </div> 901 902 903 <!-- _______________________________________________________________________ --> 904 <h4> 905 <a name="licm">The <tt>-licm</tt> pass</a> 906 </h4> 907 908 <div> 909 910 <p>The <tt>-licm</tt> pass implements various Loop Invariant Code Motion related 911 transformations. It uses the <tt>AliasAnalysis</tt> interface for several 912 different transformations:</p> 913 914 <ul> 915 <li>It uses mod/ref information to hoist or sink load instructions out of loops 916 if there are no instructions in the loop that modifies the memory loaded.</li> 917 918 <li>It uses mod/ref information to hoist function calls out of loops that do not 919 write to memory and are loop-invariant.</li> 920 921 <li>If uses alias information to promote memory objects that are loaded and 922 stored to in loops to live in a register instead. It can do this if there are 923 no may aliases to the loaded/stored memory location.</li> 924 </ul> 925 926 </div> 927 928 <!-- _______________________________________________________________________ --> 929 <h4> 930 <a name="argpromotion">The <tt>-argpromotion</tt> pass</a> 931 </h4> 932 933 <div> 934 <p> 935 The <tt>-argpromotion</tt> pass promotes by-reference arguments to be passed in 936 by-value instead. In particular, if pointer arguments are only loaded from it 937 passes in the value loaded instead of the address to the function. This pass 938 uses alias information to make sure that the value loaded from the argument 939 pointer is not modified between the entry of the function and any load of the 940 pointer.</p> 941 </div> 942 943 <!-- _______________________________________________________________________ --> 944 <h4> 945 <a name="gvn">The <tt>-gvn</tt>, <tt>-memcpyopt</tt>, and <tt>-dse</tt> 946 passes</a> 947 </h4> 948 949 <div> 950 951 <p>These passes use AliasAnalysis information to reason about loads and stores. 952 </p> 953 954 </div> 955 956 </div> 957 958 <!-- ======================================================================= --> 959 <h3> 960 <a name="aliasanalysis-debug">Clients for debugging and evaluation of 961 implementations</a> 962 </h3> 963 964 <div> 965 966 <p>These passes are useful for evaluating the various alias analysis 967 implementations. You can use them with commands like '<tt>opt -ds-aa 968 -aa-eval foo.bc -disable-output -stats</tt>'.</p> 969 970 <!-- _______________________________________________________________________ --> 971 <h4> 972 <a name="print-alias-sets">The <tt>-print-alias-sets</tt> pass</a> 973 </h4> 974 975 <div> 976 977 <p>The <tt>-print-alias-sets</tt> pass is exposed as part of the 978 <tt>opt</tt> tool to print out the Alias Sets formed by the <a 979 href="#ast"><tt>AliasSetTracker</tt></a> class. This is useful if you're using 980 the <tt>AliasSetTracker</tt> class. To use it, use something like:</p> 981 982 <div class="doc_code"> 983 <pre> 984 % opt -ds-aa -print-alias-sets -disable-output 985 </pre> 986 </div> 987 988 </div> 989 990 991 <!-- _______________________________________________________________________ --> 992 <h4> 993 <a name="count-aa">The <tt>-count-aa</tt> pass</a> 994 </h4> 995 996 <div> 997 998 <p>The <tt>-count-aa</tt> pass is useful to see how many queries a particular 999 pass is making and what responses are returned by the alias analysis. As an 1000 example,</p> 1001 1002 <div class="doc_code"> 1003 <pre> 1004 % opt -basicaa -count-aa -ds-aa -count-aa -licm 1005 </pre> 1006 </div> 1007 1008 <p>will print out how many queries (and what responses are returned) by the 1009 <tt>-licm</tt> pass (of the <tt>-ds-aa</tt> pass) and how many queries are made 1010 of the <tt>-basicaa</tt> pass by the <tt>-ds-aa</tt> pass. This can be useful 1011 when debugging a transformation or an alias analysis implementation.</p> 1012 1013 </div> 1014 1015 <!-- _______________________________________________________________________ --> 1016 <h4> 1017 <a name="aa-eval">The <tt>-aa-eval</tt> pass</a> 1018 </h4> 1019 1020 <div> 1021 1022 <p>The <tt>-aa-eval</tt> pass simply iterates through all pairs of pointers in a 1023 function and asks an alias analysis whether or not the pointers alias. This 1024 gives an indication of the precision of the alias analysis. Statistics are 1025 printed indicating the percent of no/may/must aliases found (a more precise 1026 algorithm will have a lower number of may aliases).</p> 1027 1028 </div> 1029 1030 </div> 1031 1032 </div> 1033 1034 <!-- *********************************************************************** --> 1035 <h2> 1036 <a name="memdep">Memory Dependence Analysis</a> 1037 </h2> 1038 <!-- *********************************************************************** --> 1039 1040 <div> 1041 1042 <p>If you're just looking to be a client of alias analysis information, consider 1043 using the Memory Dependence Analysis interface instead. MemDep is a lazy, 1044 caching layer on top of alias analysis that is able to answer the question of 1045 what preceding memory operations a given instruction depends on, either at an 1046 intra- or inter-block level. Because of its laziness and caching 1047 policy, using MemDep can be a significant performance win over accessing alias 1048 analysis directly.</p> 1049 1050 </div> 1051 1052 <!-- *********************************************************************** --> 1053 1054 <hr> 1055 <address> 1056 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img 1057 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a> 1058 <a href="http://validator.w3.org/check/referer"><img 1059 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a> 1060 1061 <a href="mailto:sabre (a] nondot.org">Chris Lattner</a><br> 1062 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br> 1063 Last modified: $Date: 2011-05-24 18:01:32 -0400 (Tue, 24 May 2011) $ 1064 </address> 1065 1066 </body> 1067 </html> 1068