1 | //==- AliasAnalysis.cpp - Generic Alias Analysis Interface Implementation --==// |
2 | // |
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
4 | // See https://llvm.org/LICENSE.txt for license information. |
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
6 | // |
7 | //===----------------------------------------------------------------------===// |
8 | // |
9 | // This file implements the generic AliasAnalysis interface which is used as the |
10 | // common interface used by all clients and implementations of alias analysis. |
11 | // |
12 | // This file also implements the default version of the AliasAnalysis interface |
13 | // that is to be used when no other implementation is specified. This does some |
14 | // simple tests that detect obvious cases: two different global pointers cannot |
15 | // alias, a global cannot alias a malloc, two different mallocs cannot alias, |
16 | // etc. |
17 | // |
18 | // This alias analysis implementation really isn't very good for anything, but |
19 | // it is very fast, and makes a nice clean default implementation. Because it |
20 | // handles lots of little corner cases, other, more complex, alias analysis |
21 | // implementations may choose to rely on this pass to resolve these simple and |
22 | // easy cases. |
23 | // |
24 | //===----------------------------------------------------------------------===// |
25 | |
26 | #include "llvm/Analysis/AliasAnalysis.h" |
27 | #include "llvm/ADT/Statistic.h" |
28 | #include "llvm/Analysis/BasicAliasAnalysis.h" |
29 | #include "llvm/Analysis/CaptureTracking.h" |
30 | #include "llvm/Analysis/GlobalsModRef.h" |
31 | #include "llvm/Analysis/MemoryLocation.h" |
32 | #include "llvm/Analysis/ObjCARCAliasAnalysis.h" |
33 | #include "llvm/Analysis/ScalarEvolutionAliasAnalysis.h" |
34 | #include "llvm/Analysis/ScopedNoAliasAA.h" |
35 | #include "llvm/Analysis/TargetLibraryInfo.h" |
36 | #include "llvm/Analysis/TypeBasedAliasAnalysis.h" |
37 | #include "llvm/Analysis/ValueTracking.h" |
38 | #include "llvm/IR/Argument.h" |
39 | #include "llvm/IR/Attributes.h" |
40 | #include "llvm/IR/BasicBlock.h" |
41 | #include "llvm/IR/Instruction.h" |
42 | #include "llvm/IR/Instructions.h" |
43 | #include "llvm/IR/Type.h" |
44 | #include "llvm/IR/Value.h" |
45 | #include "llvm/InitializePasses.h" |
46 | #include "llvm/Pass.h" |
47 | #include "llvm/Support/AtomicOrdering.h" |
48 | #include "llvm/Support/Casting.h" |
49 | #include "llvm/Support/CommandLine.h" |
50 | #include <algorithm> |
51 | #include <cassert> |
52 | #include <functional> |
53 | #include <iterator> |
54 | |
55 | #define DEBUG_TYPE "aa" |
56 | |
57 | using namespace llvm; |
58 | |
59 | STATISTIC(NumNoAlias, "Number of NoAlias results" ); |
60 | STATISTIC(NumMayAlias, "Number of MayAlias results" ); |
61 | STATISTIC(NumMustAlias, "Number of MustAlias results" ); |
62 | |
63 | namespace llvm { |
64 | /// Allow disabling BasicAA from the AA results. This is particularly useful |
65 | /// when testing to isolate a single AA implementation. |
66 | cl::opt<bool> DisableBasicAA("disable-basic-aa" , cl::Hidden, cl::init(Val: false)); |
67 | } // namespace llvm |
68 | |
69 | #ifndef NDEBUG |
70 | /// Print a trace of alias analysis queries and their results. |
71 | static cl::opt<bool> EnableAATrace("aa-trace" , cl::Hidden, cl::init(false)); |
72 | #else |
73 | static const bool EnableAATrace = false; |
74 | #endif |
75 | |
76 | AAResults::AAResults(AAResults &&Arg) |
77 | : TLI(Arg.TLI), AAs(std::move(Arg.AAs)), AADeps(std::move(Arg.AADeps)) {} |
78 | |
79 | AAResults::~AAResults() {} |
80 | |
81 | bool AAResults::invalidate(Function &F, const PreservedAnalyses &PA, |
82 | FunctionAnalysisManager::Invalidator &Inv) { |
83 | // AAResults preserves the AAManager by default, due to the stateless nature |
84 | // of AliasAnalysis. There is no need to check whether it has been preserved |
85 | // explicitly. Check if any module dependency was invalidated and caused the |
86 | // AAManager to be invalidated. Invalidate ourselves in that case. |
87 | auto PAC = PA.getChecker<AAManager>(); |
88 | if (!PAC.preservedWhenStateless()) |
89 | return true; |
90 | |
91 | // Check if any of the function dependencies were invalidated, and invalidate |
92 | // ourselves in that case. |
93 | for (AnalysisKey *ID : AADeps) |
94 | if (Inv.invalidate(ID, IR&: F, PA)) |
95 | return true; |
96 | |
97 | // Everything we depend on is still fine, so are we. Nothing to invalidate. |
98 | return false; |
99 | } |
100 | |
101 | //===----------------------------------------------------------------------===// |
102 | // Default chaining methods |
103 | //===----------------------------------------------------------------------===// |
104 | |
105 | AliasResult AAResults::alias(const MemoryLocation &LocA, |
106 | const MemoryLocation &LocB) { |
107 | SimpleAAQueryInfo AAQIP(*this); |
108 | return alias(LocA, LocB, AAQI&: AAQIP, CtxI: nullptr); |
109 | } |
110 | |
111 | AliasResult AAResults::alias(const MemoryLocation &LocA, |
112 | const MemoryLocation &LocB, AAQueryInfo &AAQI, |
113 | const Instruction *CtxI) { |
114 | AliasResult Result = AliasResult::MayAlias; |
115 | |
116 | if (EnableAATrace) { |
117 | for (unsigned I = 0; I < AAQI.Depth; ++I) |
118 | dbgs() << " " ; |
119 | dbgs() << "Start " << *LocA.Ptr << " @ " << LocA.Size << ", " |
120 | << *LocB.Ptr << " @ " << LocB.Size << "\n" ; |
121 | } |
122 | |
123 | AAQI.Depth++; |
124 | for (const auto &AA : AAs) { |
125 | Result = AA->alias(LocA, LocB, AAQI, CtxI); |
126 | if (Result != AliasResult::MayAlias) |
127 | break; |
128 | } |
129 | AAQI.Depth--; |
130 | |
131 | if (EnableAATrace) { |
132 | for (unsigned I = 0; I < AAQI.Depth; ++I) |
133 | dbgs() << " " ; |
134 | dbgs() << "End " << *LocA.Ptr << " @ " << LocA.Size << ", " |
135 | << *LocB.Ptr << " @ " << LocB.Size << " = " << Result << "\n" ; |
136 | } |
137 | |
138 | if (AAQI.Depth == 0) { |
139 | if (Result == AliasResult::NoAlias) |
140 | ++NumNoAlias; |
141 | else if (Result == AliasResult::MustAlias) |
142 | ++NumMustAlias; |
143 | else |
144 | ++NumMayAlias; |
145 | } |
146 | return Result; |
147 | } |
148 | |
149 | ModRefInfo AAResults::getModRefInfoMask(const MemoryLocation &Loc, |
150 | bool IgnoreLocals) { |
151 | SimpleAAQueryInfo AAQIP(*this); |
152 | return getModRefInfoMask(Loc, AAQI&: AAQIP, IgnoreLocals); |
153 | } |
154 | |
155 | ModRefInfo AAResults::getModRefInfoMask(const MemoryLocation &Loc, |
156 | AAQueryInfo &AAQI, bool IgnoreLocals) { |
157 | ModRefInfo Result = ModRefInfo::ModRef; |
158 | |
159 | for (const auto &AA : AAs) { |
160 | Result &= AA->getModRefInfoMask(Loc, AAQI, IgnoreLocals); |
161 | |
162 | // Early-exit the moment we reach the bottom of the lattice. |
163 | if (isNoModRef(MRI: Result)) |
164 | return ModRefInfo::NoModRef; |
165 | } |
166 | |
167 | return Result; |
168 | } |
169 | |
170 | ModRefInfo AAResults::getArgModRefInfo(const CallBase *Call, unsigned ArgIdx) { |
171 | ModRefInfo Result = ModRefInfo::ModRef; |
172 | |
173 | for (const auto &AA : AAs) { |
174 | Result &= AA->getArgModRefInfo(Call, ArgIdx); |
175 | |
176 | // Early-exit the moment we reach the bottom of the lattice. |
177 | if (isNoModRef(MRI: Result)) |
178 | return ModRefInfo::NoModRef; |
179 | } |
180 | |
181 | return Result; |
182 | } |
183 | |
184 | ModRefInfo AAResults::getModRefInfo(const Instruction *I, |
185 | const CallBase *Call2) { |
186 | SimpleAAQueryInfo AAQIP(*this); |
187 | return getModRefInfo(I, Call2, AAQIP); |
188 | } |
189 | |
190 | ModRefInfo AAResults::getModRefInfo(const Instruction *I, const CallBase *Call2, |
191 | AAQueryInfo &AAQI) { |
192 | // We may have two calls. |
193 | if (const auto *Call1 = dyn_cast<CallBase>(Val: I)) { |
194 | // Check if the two calls modify the same memory. |
195 | return getModRefInfo(Call1, Call2, AAQI); |
196 | } |
197 | // If this is a fence, just return ModRef. |
198 | if (I->isFenceLike()) |
199 | return ModRefInfo::ModRef; |
200 | // Otherwise, check if the call modifies or references the |
201 | // location this memory access defines. The best we can say |
202 | // is that if the call references what this instruction |
203 | // defines, it must be clobbered by this location. |
204 | const MemoryLocation DefLoc = MemoryLocation::get(Inst: I); |
205 | ModRefInfo MR = getModRefInfo(Call: Call2, Loc: DefLoc, AAQI); |
206 | if (isModOrRefSet(MRI: MR)) |
207 | return ModRefInfo::ModRef; |
208 | return ModRefInfo::NoModRef; |
209 | } |
210 | |
211 | ModRefInfo AAResults::getModRefInfo(const CallBase *Call, |
212 | const MemoryLocation &Loc, |
213 | AAQueryInfo &AAQI) { |
214 | ModRefInfo Result = ModRefInfo::ModRef; |
215 | |
216 | for (const auto &AA : AAs) { |
217 | Result &= AA->getModRefInfo(Call, Loc, AAQI); |
218 | |
219 | // Early-exit the moment we reach the bottom of the lattice. |
220 | if (isNoModRef(MRI: Result)) |
221 | return ModRefInfo::NoModRef; |
222 | } |
223 | |
224 | // Try to refine the mod-ref info further using other API entry points to the |
225 | // aggregate set of AA results. |
226 | |
227 | // We can completely ignore inaccessible memory here, because MemoryLocations |
228 | // can only reference accessible memory. |
229 | auto ME = getMemoryEffects(Call, AAQI) |
230 | .getWithoutLoc(Loc: IRMemLocation::InaccessibleMem); |
231 | if (ME.doesNotAccessMemory()) |
232 | return ModRefInfo::NoModRef; |
233 | |
234 | ModRefInfo ArgMR = ME.getModRef(Loc: IRMemLocation::ArgMem); |
235 | ModRefInfo OtherMR = ME.getWithoutLoc(Loc: IRMemLocation::ArgMem).getModRef(); |
236 | if ((ArgMR | OtherMR) != OtherMR) { |
237 | // Refine the modref info for argument memory. We only bother to do this |
238 | // if ArgMR is not a subset of OtherMR, otherwise this won't have an impact |
239 | // on the final result. |
240 | ModRefInfo AllArgsMask = ModRefInfo::NoModRef; |
241 | for (const auto &I : llvm::enumerate(First: Call->args())) { |
242 | const Value *Arg = I.value(); |
243 | if (!Arg->getType()->isPointerTy()) |
244 | continue; |
245 | unsigned ArgIdx = I.index(); |
246 | MemoryLocation ArgLoc = MemoryLocation::getForArgument(Call, ArgIdx, TLI); |
247 | AliasResult ArgAlias = alias(LocA: ArgLoc, LocB: Loc, AAQI, CtxI: Call); |
248 | if (ArgAlias != AliasResult::NoAlias) |
249 | AllArgsMask |= getArgModRefInfo(Call, ArgIdx); |
250 | } |
251 | ArgMR &= AllArgsMask; |
252 | } |
253 | |
254 | Result &= ArgMR | OtherMR; |
255 | |
256 | // Apply the ModRef mask. This ensures that if Loc is a constant memory |
257 | // location, we take into account the fact that the call definitely could not |
258 | // modify the memory location. |
259 | if (!isNoModRef(MRI: Result)) |
260 | Result &= getModRefInfoMask(Loc); |
261 | |
262 | return Result; |
263 | } |
264 | |
265 | ModRefInfo AAResults::getModRefInfo(const CallBase *Call1, |
266 | const CallBase *Call2, AAQueryInfo &AAQI) { |
267 | ModRefInfo Result = ModRefInfo::ModRef; |
268 | |
269 | for (const auto &AA : AAs) { |
270 | Result &= AA->getModRefInfo(Call1, Call2, AAQI); |
271 | |
272 | // Early-exit the moment we reach the bottom of the lattice. |
273 | if (isNoModRef(MRI: Result)) |
274 | return ModRefInfo::NoModRef; |
275 | } |
276 | |
277 | // Try to refine the mod-ref info further using other API entry points to the |
278 | // aggregate set of AA results. |
279 | |
280 | // If Call1 or Call2 are readnone, they don't interact. |
281 | auto Call1B = getMemoryEffects(Call: Call1, AAQI); |
282 | if (Call1B.doesNotAccessMemory()) |
283 | return ModRefInfo::NoModRef; |
284 | |
285 | auto Call2B = getMemoryEffects(Call: Call2, AAQI); |
286 | if (Call2B.doesNotAccessMemory()) |
287 | return ModRefInfo::NoModRef; |
288 | |
289 | // If they both only read from memory, there is no dependence. |
290 | if (Call1B.onlyReadsMemory() && Call2B.onlyReadsMemory()) |
291 | return ModRefInfo::NoModRef; |
292 | |
293 | // If Call1 only reads memory, the only dependence on Call2 can be |
294 | // from Call1 reading memory written by Call2. |
295 | if (Call1B.onlyReadsMemory()) |
296 | Result &= ModRefInfo::Ref; |
297 | else if (Call1B.onlyWritesMemory()) |
298 | Result &= ModRefInfo::Mod; |
299 | |
300 | // If Call2 only access memory through arguments, accumulate the mod/ref |
301 | // information from Call1's references to the memory referenced by |
302 | // Call2's arguments. |
303 | if (Call2B.onlyAccessesArgPointees()) { |
304 | if (!Call2B.doesAccessArgPointees()) |
305 | return ModRefInfo::NoModRef; |
306 | ModRefInfo R = ModRefInfo::NoModRef; |
307 | for (auto I = Call2->arg_begin(), E = Call2->arg_end(); I != E; ++I) { |
308 | const Value *Arg = *I; |
309 | if (!Arg->getType()->isPointerTy()) |
310 | continue; |
311 | unsigned Call2ArgIdx = std::distance(first: Call2->arg_begin(), last: I); |
312 | auto Call2ArgLoc = |
313 | MemoryLocation::getForArgument(Call: Call2, ArgIdx: Call2ArgIdx, TLI); |
314 | |
315 | // ArgModRefC2 indicates what Call2 might do to Call2ArgLoc, and the |
316 | // dependence of Call1 on that location is the inverse: |
317 | // - If Call2 modifies location, dependence exists if Call1 reads or |
318 | // writes. |
319 | // - If Call2 only reads location, dependence exists if Call1 writes. |
320 | ModRefInfo ArgModRefC2 = getArgModRefInfo(Call: Call2, ArgIdx: Call2ArgIdx); |
321 | ModRefInfo ArgMask = ModRefInfo::NoModRef; |
322 | if (isModSet(MRI: ArgModRefC2)) |
323 | ArgMask = ModRefInfo::ModRef; |
324 | else if (isRefSet(MRI: ArgModRefC2)) |
325 | ArgMask = ModRefInfo::Mod; |
326 | |
327 | // ModRefC1 indicates what Call1 might do to Call2ArgLoc, and we use |
328 | // above ArgMask to update dependence info. |
329 | ArgMask &= getModRefInfo(Call: Call1, Loc: Call2ArgLoc, AAQI); |
330 | |
331 | R = (R | ArgMask) & Result; |
332 | if (R == Result) |
333 | break; |
334 | } |
335 | |
336 | return R; |
337 | } |
338 | |
339 | // If Call1 only accesses memory through arguments, check if Call2 references |
340 | // any of the memory referenced by Call1's arguments. If not, return NoModRef. |
341 | if (Call1B.onlyAccessesArgPointees()) { |
342 | if (!Call1B.doesAccessArgPointees()) |
343 | return ModRefInfo::NoModRef; |
344 | ModRefInfo R = ModRefInfo::NoModRef; |
345 | for (auto I = Call1->arg_begin(), E = Call1->arg_end(); I != E; ++I) { |
346 | const Value *Arg = *I; |
347 | if (!Arg->getType()->isPointerTy()) |
348 | continue; |
349 | unsigned Call1ArgIdx = std::distance(first: Call1->arg_begin(), last: I); |
350 | auto Call1ArgLoc = |
351 | MemoryLocation::getForArgument(Call: Call1, ArgIdx: Call1ArgIdx, TLI); |
352 | |
353 | // ArgModRefC1 indicates what Call1 might do to Call1ArgLoc; if Call1 |
354 | // might Mod Call1ArgLoc, then we care about either a Mod or a Ref by |
355 | // Call2. If Call1 might Ref, then we care only about a Mod by Call2. |
356 | ModRefInfo ArgModRefC1 = getArgModRefInfo(Call: Call1, ArgIdx: Call1ArgIdx); |
357 | ModRefInfo ModRefC2 = getModRefInfo(Call: Call2, Loc: Call1ArgLoc, AAQI); |
358 | if ((isModSet(MRI: ArgModRefC1) && isModOrRefSet(MRI: ModRefC2)) || |
359 | (isRefSet(MRI: ArgModRefC1) && isModSet(MRI: ModRefC2))) |
360 | R = (R | ArgModRefC1) & Result; |
361 | |
362 | if (R == Result) |
363 | break; |
364 | } |
365 | |
366 | return R; |
367 | } |
368 | |
369 | return Result; |
370 | } |
371 | |
372 | MemoryEffects AAResults::getMemoryEffects(const CallBase *Call, |
373 | AAQueryInfo &AAQI) { |
374 | MemoryEffects Result = MemoryEffects::unknown(); |
375 | |
376 | for (const auto &AA : AAs) { |
377 | Result &= AA->getMemoryEffects(Call, AAQI); |
378 | |
379 | // Early-exit the moment we reach the bottom of the lattice. |
380 | if (Result.doesNotAccessMemory()) |
381 | return Result; |
382 | } |
383 | |
384 | return Result; |
385 | } |
386 | |
387 | MemoryEffects AAResults::getMemoryEffects(const CallBase *Call) { |
388 | SimpleAAQueryInfo AAQI(*this); |
389 | return getMemoryEffects(Call, AAQI); |
390 | } |
391 | |
392 | MemoryEffects AAResults::getMemoryEffects(const Function *F) { |
393 | MemoryEffects Result = MemoryEffects::unknown(); |
394 | |
395 | for (const auto &AA : AAs) { |
396 | Result &= AA->getMemoryEffects(F); |
397 | |
398 | // Early-exit the moment we reach the bottom of the lattice. |
399 | if (Result.doesNotAccessMemory()) |
400 | return Result; |
401 | } |
402 | |
403 | return Result; |
404 | } |
405 | |
406 | raw_ostream &llvm::operator<<(raw_ostream &OS, AliasResult AR) { |
407 | switch (AR) { |
408 | case AliasResult::NoAlias: |
409 | OS << "NoAlias" ; |
410 | break; |
411 | case AliasResult::MustAlias: |
412 | OS << "MustAlias" ; |
413 | break; |
414 | case AliasResult::MayAlias: |
415 | OS << "MayAlias" ; |
416 | break; |
417 | case AliasResult::PartialAlias: |
418 | OS << "PartialAlias" ; |
419 | if (AR.hasOffset()) |
420 | OS << " (off " << AR.getOffset() << ")" ; |
421 | break; |
422 | } |
423 | return OS; |
424 | } |
425 | |
426 | raw_ostream &llvm::operator<<(raw_ostream &OS, ModRefInfo MR) { |
427 | switch (MR) { |
428 | case ModRefInfo::NoModRef: |
429 | OS << "NoModRef" ; |
430 | break; |
431 | case ModRefInfo::Ref: |
432 | OS << "Ref" ; |
433 | break; |
434 | case ModRefInfo::Mod: |
435 | OS << "Mod" ; |
436 | break; |
437 | case ModRefInfo::ModRef: |
438 | OS << "ModRef" ; |
439 | break; |
440 | } |
441 | return OS; |
442 | } |
443 | |
444 | raw_ostream &llvm::operator<<(raw_ostream &OS, MemoryEffects ME) { |
445 | for (IRMemLocation Loc : MemoryEffects::locations()) { |
446 | switch (Loc) { |
447 | case IRMemLocation::ArgMem: |
448 | OS << "ArgMem: " ; |
449 | break; |
450 | case IRMemLocation::InaccessibleMem: |
451 | OS << "InaccessibleMem: " ; |
452 | break; |
453 | case IRMemLocation::Other: |
454 | OS << "Other: " ; |
455 | break; |
456 | } |
457 | OS << ME.getModRef(Loc) << ", " ; |
458 | } |
459 | return OS; |
460 | } |
461 | |
462 | //===----------------------------------------------------------------------===// |
463 | // Helper method implementation |
464 | //===----------------------------------------------------------------------===// |
465 | |
466 | ModRefInfo AAResults::getModRefInfo(const LoadInst *L, |
467 | const MemoryLocation &Loc, |
468 | AAQueryInfo &AAQI) { |
469 | // Be conservative in the face of atomic. |
470 | if (isStrongerThan(AO: L->getOrdering(), Other: AtomicOrdering::Unordered)) |
471 | return ModRefInfo::ModRef; |
472 | |
473 | // If the load address doesn't alias the given address, it doesn't read |
474 | // or write the specified memory. |
475 | if (Loc.Ptr) { |
476 | AliasResult AR = alias(LocA: MemoryLocation::get(LI: L), LocB: Loc, AAQI, CtxI: L); |
477 | if (AR == AliasResult::NoAlias) |
478 | return ModRefInfo::NoModRef; |
479 | } |
480 | // Otherwise, a load just reads. |
481 | return ModRefInfo::Ref; |
482 | } |
483 | |
484 | ModRefInfo AAResults::getModRefInfo(const StoreInst *S, |
485 | const MemoryLocation &Loc, |
486 | AAQueryInfo &AAQI) { |
487 | // Be conservative in the face of atomic. |
488 | if (isStrongerThan(AO: S->getOrdering(), Other: AtomicOrdering::Unordered)) |
489 | return ModRefInfo::ModRef; |
490 | |
491 | if (Loc.Ptr) { |
492 | AliasResult AR = alias(LocA: MemoryLocation::get(SI: S), LocB: Loc, AAQI, CtxI: S); |
493 | // If the store address cannot alias the pointer in question, then the |
494 | // specified memory cannot be modified by the store. |
495 | if (AR == AliasResult::NoAlias) |
496 | return ModRefInfo::NoModRef; |
497 | |
498 | // Examine the ModRef mask. If Mod isn't present, then return NoModRef. |
499 | // This ensures that if Loc is a constant memory location, we take into |
500 | // account the fact that the store definitely could not modify the memory |
501 | // location. |
502 | if (!isModSet(MRI: getModRefInfoMask(Loc))) |
503 | return ModRefInfo::NoModRef; |
504 | } |
505 | |
506 | // Otherwise, a store just writes. |
507 | return ModRefInfo::Mod; |
508 | } |
509 | |
510 | ModRefInfo AAResults::getModRefInfo(const FenceInst *S, |
511 | const MemoryLocation &Loc, |
512 | AAQueryInfo &AAQI) { |
513 | // All we know about a fence instruction is what we get from the ModRef |
514 | // mask: if Loc is a constant memory location, the fence definitely could |
515 | // not modify it. |
516 | if (Loc.Ptr) |
517 | return getModRefInfoMask(Loc); |
518 | return ModRefInfo::ModRef; |
519 | } |
520 | |
521 | ModRefInfo AAResults::getModRefInfo(const VAArgInst *V, |
522 | const MemoryLocation &Loc, |
523 | AAQueryInfo &AAQI) { |
524 | if (Loc.Ptr) { |
525 | AliasResult AR = alias(LocA: MemoryLocation::get(VI: V), LocB: Loc, AAQI, CtxI: V); |
526 | // If the va_arg address cannot alias the pointer in question, then the |
527 | // specified memory cannot be accessed by the va_arg. |
528 | if (AR == AliasResult::NoAlias) |
529 | return ModRefInfo::NoModRef; |
530 | |
531 | // If the pointer is a pointer to invariant memory, then it could not have |
532 | // been modified by this va_arg. |
533 | return getModRefInfoMask(Loc, AAQI); |
534 | } |
535 | |
536 | // Otherwise, a va_arg reads and writes. |
537 | return ModRefInfo::ModRef; |
538 | } |
539 | |
540 | ModRefInfo AAResults::getModRefInfo(const CatchPadInst *CatchPad, |
541 | const MemoryLocation &Loc, |
542 | AAQueryInfo &AAQI) { |
543 | if (Loc.Ptr) { |
544 | // If the pointer is a pointer to invariant memory, |
545 | // then it could not have been modified by this catchpad. |
546 | return getModRefInfoMask(Loc, AAQI); |
547 | } |
548 | |
549 | // Otherwise, a catchpad reads and writes. |
550 | return ModRefInfo::ModRef; |
551 | } |
552 | |
553 | ModRefInfo AAResults::getModRefInfo(const CatchReturnInst *CatchRet, |
554 | const MemoryLocation &Loc, |
555 | AAQueryInfo &AAQI) { |
556 | if (Loc.Ptr) { |
557 | // If the pointer is a pointer to invariant memory, |
558 | // then it could not have been modified by this catchpad. |
559 | return getModRefInfoMask(Loc, AAQI); |
560 | } |
561 | |
562 | // Otherwise, a catchret reads and writes. |
563 | return ModRefInfo::ModRef; |
564 | } |
565 | |
566 | ModRefInfo AAResults::getModRefInfo(const AtomicCmpXchgInst *CX, |
567 | const MemoryLocation &Loc, |
568 | AAQueryInfo &AAQI) { |
569 | // Acquire/Release cmpxchg has properties that matter for arbitrary addresses. |
570 | if (isStrongerThanMonotonic(AO: CX->getSuccessOrdering())) |
571 | return ModRefInfo::ModRef; |
572 | |
573 | if (Loc.Ptr) { |
574 | AliasResult AR = alias(LocA: MemoryLocation::get(CXI: CX), LocB: Loc, AAQI, CtxI: CX); |
575 | // If the cmpxchg address does not alias the location, it does not access |
576 | // it. |
577 | if (AR == AliasResult::NoAlias) |
578 | return ModRefInfo::NoModRef; |
579 | } |
580 | |
581 | return ModRefInfo::ModRef; |
582 | } |
583 | |
584 | ModRefInfo AAResults::getModRefInfo(const AtomicRMWInst *RMW, |
585 | const MemoryLocation &Loc, |
586 | AAQueryInfo &AAQI) { |
587 | // Acquire/Release atomicrmw has properties that matter for arbitrary addresses. |
588 | if (isStrongerThanMonotonic(AO: RMW->getOrdering())) |
589 | return ModRefInfo::ModRef; |
590 | |
591 | if (Loc.Ptr) { |
592 | AliasResult AR = alias(LocA: MemoryLocation::get(RMWI: RMW), LocB: Loc, AAQI, CtxI: RMW); |
593 | // If the atomicrmw address does not alias the location, it does not access |
594 | // it. |
595 | if (AR == AliasResult::NoAlias) |
596 | return ModRefInfo::NoModRef; |
597 | } |
598 | |
599 | return ModRefInfo::ModRef; |
600 | } |
601 | |
602 | ModRefInfo AAResults::getModRefInfo(const Instruction *I, |
603 | const std::optional<MemoryLocation> &OptLoc, |
604 | AAQueryInfo &AAQIP) { |
605 | if (OptLoc == std::nullopt) { |
606 | if (const auto *Call = dyn_cast<CallBase>(Val: I)) |
607 | return getMemoryEffects(Call, AAQI&: AAQIP).getModRef(); |
608 | } |
609 | |
610 | const MemoryLocation &Loc = OptLoc.value_or(u: MemoryLocation()); |
611 | |
612 | switch (I->getOpcode()) { |
613 | case Instruction::VAArg: |
614 | return getModRefInfo(V: (const VAArgInst *)I, Loc, AAQI&: AAQIP); |
615 | case Instruction::Load: |
616 | return getModRefInfo(L: (const LoadInst *)I, Loc, AAQI&: AAQIP); |
617 | case Instruction::Store: |
618 | return getModRefInfo(S: (const StoreInst *)I, Loc, AAQI&: AAQIP); |
619 | case Instruction::Fence: |
620 | return getModRefInfo(S: (const FenceInst *)I, Loc, AAQI&: AAQIP); |
621 | case Instruction::AtomicCmpXchg: |
622 | return getModRefInfo(CX: (const AtomicCmpXchgInst *)I, Loc, AAQI&: AAQIP); |
623 | case Instruction::AtomicRMW: |
624 | return getModRefInfo(RMW: (const AtomicRMWInst *)I, Loc, AAQI&: AAQIP); |
625 | case Instruction::Call: |
626 | case Instruction::CallBr: |
627 | case Instruction::Invoke: |
628 | return getModRefInfo(Call: (const CallBase *)I, Loc, AAQI&: AAQIP); |
629 | case Instruction::CatchPad: |
630 | return getModRefInfo(CatchPad: (const CatchPadInst *)I, Loc, AAQI&: AAQIP); |
631 | case Instruction::CatchRet: |
632 | return getModRefInfo(CatchRet: (const CatchReturnInst *)I, Loc, AAQI&: AAQIP); |
633 | default: |
634 | assert(!I->mayReadOrWriteMemory() && |
635 | "Unhandled memory access instruction!" ); |
636 | return ModRefInfo::NoModRef; |
637 | } |
638 | } |
639 | |
640 | /// Return information about whether a particular call site modifies |
641 | /// or reads the specified memory location \p MemLoc before instruction \p I |
642 | /// in a BasicBlock. |
643 | /// FIXME: this is really just shoring-up a deficiency in alias analysis. |
644 | /// BasicAA isn't willing to spend linear time determining whether an alloca |
645 | /// was captured before or after this particular call, while we are. However, |
646 | /// with a smarter AA in place, this test is just wasting compile time. |
647 | ModRefInfo AAResults::callCapturesBefore(const Instruction *I, |
648 | const MemoryLocation &MemLoc, |
649 | DominatorTree *DT, |
650 | AAQueryInfo &AAQI) { |
651 | if (!DT) |
652 | return ModRefInfo::ModRef; |
653 | |
654 | const Value *Object = getUnderlyingObject(V: MemLoc.Ptr); |
655 | if (!isIdentifiedFunctionLocal(V: Object)) |
656 | return ModRefInfo::ModRef; |
657 | |
658 | const auto *Call = dyn_cast<CallBase>(Val: I); |
659 | if (!Call || Call == Object) |
660 | return ModRefInfo::ModRef; |
661 | |
662 | if (PointerMayBeCapturedBefore(V: Object, /* ReturnCaptures */ true, |
663 | /* StoreCaptures */ true, I, DT, |
664 | /* include Object */ IncludeI: true)) |
665 | return ModRefInfo::ModRef; |
666 | |
667 | unsigned ArgNo = 0; |
668 | ModRefInfo R = ModRefInfo::NoModRef; |
669 | // Set flag only if no May found and all operands processed. |
670 | for (auto CI = Call->data_operands_begin(), CE = Call->data_operands_end(); |
671 | CI != CE; ++CI, ++ArgNo) { |
672 | // Only look at the no-capture or byval pointer arguments. If this |
673 | // pointer were passed to arguments that were neither of these, then it |
674 | // couldn't be no-capture. |
675 | if (!(*CI)->getType()->isPointerTy() || |
676 | (!Call->doesNotCapture(OpNo: ArgNo) && ArgNo < Call->arg_size() && |
677 | !Call->isByValArgument(ArgNo))) |
678 | continue; |
679 | |
680 | AliasResult AR = |
681 | alias(LocA: MemoryLocation::getBeforeOrAfter(Ptr: *CI), |
682 | LocB: MemoryLocation::getBeforeOrAfter(Ptr: Object), AAQI, CtxI: Call); |
683 | // If this is a no-capture pointer argument, see if we can tell that it |
684 | // is impossible to alias the pointer we're checking. If not, we have to |
685 | // assume that the call could touch the pointer, even though it doesn't |
686 | // escape. |
687 | if (AR == AliasResult::NoAlias) |
688 | continue; |
689 | if (Call->doesNotAccessMemory(OpNo: ArgNo)) |
690 | continue; |
691 | if (Call->onlyReadsMemory(OpNo: ArgNo)) { |
692 | R = ModRefInfo::Ref; |
693 | continue; |
694 | } |
695 | return ModRefInfo::ModRef; |
696 | } |
697 | return R; |
698 | } |
699 | |
700 | /// canBasicBlockModify - Return true if it is possible for execution of the |
701 | /// specified basic block to modify the location Loc. |
702 | /// |
703 | bool AAResults::canBasicBlockModify(const BasicBlock &BB, |
704 | const MemoryLocation &Loc) { |
705 | return canInstructionRangeModRef(I1: BB.front(), I2: BB.back(), Loc, Mode: ModRefInfo::Mod); |
706 | } |
707 | |
708 | /// canInstructionRangeModRef - Return true if it is possible for the |
709 | /// execution of the specified instructions to mod\ref (according to the |
710 | /// mode) the location Loc. The instructions to consider are all |
711 | /// of the instructions in the range of [I1,I2] INCLUSIVE. |
712 | /// I1 and I2 must be in the same basic block. |
713 | bool AAResults::canInstructionRangeModRef(const Instruction &I1, |
714 | const Instruction &I2, |
715 | const MemoryLocation &Loc, |
716 | const ModRefInfo Mode) { |
717 | assert(I1.getParent() == I2.getParent() && |
718 | "Instructions not in same basic block!" ); |
719 | BasicBlock::const_iterator I = I1.getIterator(); |
720 | BasicBlock::const_iterator E = I2.getIterator(); |
721 | ++E; // Convert from inclusive to exclusive range. |
722 | |
723 | for (; I != E; ++I) // Check every instruction in range |
724 | if (isModOrRefSet(MRI: getModRefInfo(I: &*I, OptLoc: Loc) & Mode)) |
725 | return true; |
726 | return false; |
727 | } |
728 | |
729 | // Provide a definition for the root virtual destructor. |
730 | AAResults::Concept::~Concept() = default; |
731 | |
732 | // Provide a definition for the static object used to identify passes. |
733 | AnalysisKey AAManager::Key; |
734 | |
735 | ExternalAAWrapperPass::ExternalAAWrapperPass() : ImmutablePass(ID) { |
736 | initializeExternalAAWrapperPassPass(*PassRegistry::getPassRegistry()); |
737 | } |
738 | |
739 | ExternalAAWrapperPass::ExternalAAWrapperPass(CallbackT CB) |
740 | : ImmutablePass(ID), CB(std::move(CB)) { |
741 | initializeExternalAAWrapperPassPass(*PassRegistry::getPassRegistry()); |
742 | } |
743 | |
744 | char ExternalAAWrapperPass::ID = 0; |
745 | |
746 | INITIALIZE_PASS(ExternalAAWrapperPass, "external-aa" , "External Alias Analysis" , |
747 | false, true) |
748 | |
749 | ImmutablePass * |
750 | llvm::createExternalAAWrapperPass(ExternalAAWrapperPass::CallbackT Callback) { |
751 | return new ExternalAAWrapperPass(std::move(Callback)); |
752 | } |
753 | |
754 | AAResultsWrapperPass::AAResultsWrapperPass() : FunctionPass(ID) { |
755 | initializeAAResultsWrapperPassPass(*PassRegistry::getPassRegistry()); |
756 | } |
757 | |
758 | char AAResultsWrapperPass::ID = 0; |
759 | |
760 | INITIALIZE_PASS_BEGIN(AAResultsWrapperPass, "aa" , |
761 | "Function Alias Analysis Results" , false, true) |
762 | INITIALIZE_PASS_DEPENDENCY(BasicAAWrapperPass) |
763 | INITIALIZE_PASS_DEPENDENCY(ExternalAAWrapperPass) |
764 | INITIALIZE_PASS_DEPENDENCY(GlobalsAAWrapperPass) |
765 | INITIALIZE_PASS_DEPENDENCY(SCEVAAWrapperPass) |
766 | INITIALIZE_PASS_DEPENDENCY(ScopedNoAliasAAWrapperPass) |
767 | INITIALIZE_PASS_DEPENDENCY(TypeBasedAAWrapperPass) |
768 | INITIALIZE_PASS_END(AAResultsWrapperPass, "aa" , |
769 | "Function Alias Analysis Results" , false, true) |
770 | |
771 | /// Run the wrapper pass to rebuild an aggregation over known AA passes. |
772 | /// |
773 | /// This is the legacy pass manager's interface to the new-style AA results |
774 | /// aggregation object. Because this is somewhat shoe-horned into the legacy |
775 | /// pass manager, we hard code all the specific alias analyses available into |
776 | /// it. While the particular set enabled is configured via commandline flags, |
777 | /// adding a new alias analysis to LLVM will require adding support for it to |
778 | /// this list. |
779 | bool AAResultsWrapperPass::runOnFunction(Function &F) { |
780 | // NB! This *must* be reset before adding new AA results to the new |
781 | // AAResults object because in the legacy pass manager, each instance |
782 | // of these will refer to the *same* immutable analyses, registering and |
783 | // unregistering themselves with them. We need to carefully tear down the |
784 | // previous object first, in this case replacing it with an empty one, before |
785 | // registering new results. |
786 | AAR.reset( |
787 | p: new AAResults(getAnalysis<TargetLibraryInfoWrapperPass>().getTLI(F))); |
788 | |
789 | // BasicAA is always available for function analyses. Also, we add it first |
790 | // so that it can trump TBAA results when it proves MustAlias. |
791 | // FIXME: TBAA should have an explicit mode to support this and then we |
792 | // should reconsider the ordering here. |
793 | if (!DisableBasicAA) |
794 | AAR->addAAResult(AAResult&: getAnalysis<BasicAAWrapperPass>().getResult()); |
795 | |
796 | // Populate the results with the currently available AAs. |
797 | if (auto *WrapperPass = getAnalysisIfAvailable<ScopedNoAliasAAWrapperPass>()) |
798 | AAR->addAAResult(AAResult&: WrapperPass->getResult()); |
799 | if (auto *WrapperPass = getAnalysisIfAvailable<TypeBasedAAWrapperPass>()) |
800 | AAR->addAAResult(AAResult&: WrapperPass->getResult()); |
801 | if (auto *WrapperPass = getAnalysisIfAvailable<GlobalsAAWrapperPass>()) |
802 | AAR->addAAResult(AAResult&: WrapperPass->getResult()); |
803 | if (auto *WrapperPass = getAnalysisIfAvailable<SCEVAAWrapperPass>()) |
804 | AAR->addAAResult(AAResult&: WrapperPass->getResult()); |
805 | |
806 | // If available, run an external AA providing callback over the results as |
807 | // well. |
808 | if (auto *WrapperPass = getAnalysisIfAvailable<ExternalAAWrapperPass>()) |
809 | if (WrapperPass->CB) |
810 | WrapperPass->CB(*this, F, *AAR); |
811 | |
812 | // Analyses don't mutate the IR, so return false. |
813 | return false; |
814 | } |
815 | |
816 | void AAResultsWrapperPass::getAnalysisUsage(AnalysisUsage &AU) const { |
817 | AU.setPreservesAll(); |
818 | AU.addRequiredTransitive<BasicAAWrapperPass>(); |
819 | AU.addRequiredTransitive<TargetLibraryInfoWrapperPass>(); |
820 | |
821 | // We also need to mark all the alias analysis passes we will potentially |
822 | // probe in runOnFunction as used here to ensure the legacy pass manager |
823 | // preserves them. This hard coding of lists of alias analyses is specific to |
824 | // the legacy pass manager. |
825 | AU.addUsedIfAvailable<ScopedNoAliasAAWrapperPass>(); |
826 | AU.addUsedIfAvailable<TypeBasedAAWrapperPass>(); |
827 | AU.addUsedIfAvailable<GlobalsAAWrapperPass>(); |
828 | AU.addUsedIfAvailable<SCEVAAWrapperPass>(); |
829 | AU.addUsedIfAvailable<ExternalAAWrapperPass>(); |
830 | } |
831 | |
832 | AAManager::Result AAManager::run(Function &F, FunctionAnalysisManager &AM) { |
833 | Result R(AM.getResult<TargetLibraryAnalysis>(IR&: F)); |
834 | for (auto &Getter : ResultGetters) |
835 | (*Getter)(F, AM, R); |
836 | return R; |
837 | } |
838 | |
839 | bool llvm::isNoAliasCall(const Value *V) { |
840 | if (const auto *Call = dyn_cast<CallBase>(Val: V)) |
841 | return Call->hasRetAttr(Kind: Attribute::NoAlias); |
842 | return false; |
843 | } |
844 | |
845 | static bool isNoAliasOrByValArgument(const Value *V) { |
846 | if (const Argument *A = dyn_cast<Argument>(Val: V)) |
847 | return A->hasNoAliasAttr() || A->hasByValAttr(); |
848 | return false; |
849 | } |
850 | |
851 | bool llvm::isIdentifiedObject(const Value *V) { |
852 | if (isa<AllocaInst>(Val: V)) |
853 | return true; |
854 | if (isa<GlobalValue>(Val: V) && !isa<GlobalAlias>(Val: V)) |
855 | return true; |
856 | if (isNoAliasCall(V)) |
857 | return true; |
858 | if (isNoAliasOrByValArgument(V)) |
859 | return true; |
860 | return false; |
861 | } |
862 | |
863 | bool llvm::isIdentifiedFunctionLocal(const Value *V) { |
864 | return isa<AllocaInst>(Val: V) || isNoAliasCall(V) || isNoAliasOrByValArgument(V); |
865 | } |
866 | |
867 | bool llvm::isEscapeSource(const Value *V) { |
868 | if (auto *CB = dyn_cast<CallBase>(Val: V)) |
869 | return !isIntrinsicReturningPointerAliasingArgumentWithoutCapturing(Call: CB, |
870 | MustPreserveNullness: true); |
871 | |
872 | // The load case works because isNonEscapingLocalObject considers all |
873 | // stores to be escapes (it passes true for the StoreCaptures argument |
874 | // to PointerMayBeCaptured). |
875 | if (isa<LoadInst>(Val: V)) |
876 | return true; |
877 | |
878 | // The inttoptr case works because isNonEscapingLocalObject considers all |
879 | // means of converting or equating a pointer to an int (ptrtoint, ptr store |
880 | // which could be followed by an integer load, ptr<->int compare) as |
881 | // escaping, and objects located at well-known addresses via platform-specific |
882 | // means cannot be considered non-escaping local objects. |
883 | if (isa<IntToPtrInst>(Val: V)) |
884 | return true; |
885 | |
886 | // Same for inttoptr constant expressions. |
887 | if (auto *CE = dyn_cast<ConstantExpr>(Val: V)) |
888 | if (CE->getOpcode() == Instruction::IntToPtr) |
889 | return true; |
890 | |
891 | return false; |
892 | } |
893 | |
894 | bool llvm::isNotVisibleOnUnwind(const Value *Object, |
895 | bool &RequiresNoCaptureBeforeUnwind) { |
896 | RequiresNoCaptureBeforeUnwind = false; |
897 | |
898 | // Alloca goes out of scope on unwind. |
899 | if (isa<AllocaInst>(Val: Object)) |
900 | return true; |
901 | |
902 | // Byval goes out of scope on unwind. |
903 | if (auto *A = dyn_cast<Argument>(Val: Object)) |
904 | return A->hasByValAttr() || A->hasAttribute(Kind: Attribute::DeadOnUnwind); |
905 | |
906 | // A noalias return is not accessible from any other code. If the pointer |
907 | // does not escape prior to the unwind, then the caller cannot access the |
908 | // memory either. |
909 | if (isNoAliasCall(V: Object)) { |
910 | RequiresNoCaptureBeforeUnwind = true; |
911 | return true; |
912 | } |
913 | |
914 | return false; |
915 | } |
916 | |
917 | // We don't consider globals as writable: While the physical memory is writable, |
918 | // we may not have provenance to perform the write. |
919 | bool llvm::isWritableObject(const Value *Object, |
920 | bool &ExplicitlyDereferenceableOnly) { |
921 | ExplicitlyDereferenceableOnly = false; |
922 | |
923 | // TODO: Alloca might not be writable after its lifetime ends. |
924 | // See https://github.com/llvm/llvm-project/issues/51838. |
925 | if (isa<AllocaInst>(Val: Object)) |
926 | return true; |
927 | |
928 | if (auto *A = dyn_cast<Argument>(Val: Object)) { |
929 | if (A->hasAttribute(Kind: Attribute::Writable)) { |
930 | ExplicitlyDereferenceableOnly = true; |
931 | return true; |
932 | } |
933 | |
934 | return A->hasByValAttr(); |
935 | } |
936 | |
937 | // TODO: Noalias shouldn't imply writability, this should check for an |
938 | // allocator function instead. |
939 | return isNoAliasCall(V: Object); |
940 | } |
941 | |