1 //===-- MemorySSAUpdater.cpp - Memory SSA Updater--------------------===//
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 MemorySSAUpdater class.
10 //
11 //===----------------------------------------------------------------===//
12 #include "llvm/Analysis/MemorySSAUpdater.h"
13 #include "llvm/Analysis/LoopIterator.h"
14 #include "llvm/ADT/STLExtras.h"
15 #include "llvm/ADT/SetVector.h"
16 #include "llvm/ADT/SmallPtrSet.h"
17 #include "llvm/Analysis/IteratedDominanceFrontier.h"
18 #include "llvm/Analysis/MemorySSA.h"
19 #include "llvm/IR/BasicBlock.h"
20 #include "llvm/IR/DataLayout.h"
21 #include "llvm/IR/Dominators.h"
22 #include "llvm/IR/GlobalVariable.h"
23 #include "llvm/IR/IRBuilder.h"
24 #include "llvm/IR/LLVMContext.h"
25 #include "llvm/IR/Metadata.h"
26 #include "llvm/IR/Module.h"
27 #include "llvm/Support/Debug.h"
28 #include "llvm/Support/FormattedStream.h"
29 #include <algorithm>
30
31 #define DEBUG_TYPE "memoryssa"
32 using namespace llvm;
33
34 // This is the marker algorithm from "Simple and Efficient Construction of
35 // Static Single Assignment Form"
36 // The simple, non-marker algorithm places phi nodes at any join
37 // Here, we place markers, and only place phi nodes if they end up necessary.
38 // They are only necessary if they break a cycle (IE we recursively visit
39 // ourselves again), or we discover, while getting the value of the operands,
40 // that there are two or more definitions needing to be merged.
41 // This still will leave non-minimal form in the case of irreducible control
42 // flow, where phi nodes may be in cycles with themselves, but unnecessary.
getPreviousDefRecursive(BasicBlock * BB,DenseMap<BasicBlock *,TrackingVH<MemoryAccess>> & CachedPreviousDef)43 MemoryAccess *MemorySSAUpdater::getPreviousDefRecursive(
44 BasicBlock *BB,
45 DenseMap<BasicBlock *, TrackingVH<MemoryAccess>> &CachedPreviousDef) {
46 // First, do a cache lookup. Without this cache, certain CFG structures
47 // (like a series of if statements) take exponential time to visit.
48 auto Cached = CachedPreviousDef.find(BB);
49 if (Cached != CachedPreviousDef.end())
50 return Cached->second;
51
52 // If this method is called from an unreachable block, return LoE.
53 if (!MSSA->DT->isReachableFromEntry(BB))
54 return MSSA->getLiveOnEntryDef();
55
56 if (BasicBlock *Pred = BB->getUniquePredecessor()) {
57 VisitedBlocks.insert(BB);
58 // Single predecessor case, just recurse, we can only have one definition.
59 MemoryAccess *Result = getPreviousDefFromEnd(Pred, CachedPreviousDef);
60 CachedPreviousDef.insert({BB, Result});
61 return Result;
62 }
63
64 if (VisitedBlocks.count(BB)) {
65 // We hit our node again, meaning we had a cycle, we must insert a phi
66 // node to break it so we have an operand. The only case this will
67 // insert useless phis is if we have irreducible control flow.
68 MemoryAccess *Result = MSSA->createMemoryPhi(BB);
69 CachedPreviousDef.insert({BB, Result});
70 return Result;
71 }
72
73 if (VisitedBlocks.insert(BB).second) {
74 // Mark us visited so we can detect a cycle
75 SmallVector<TrackingVH<MemoryAccess>, 8> PhiOps;
76
77 // Recurse to get the values in our predecessors for placement of a
78 // potential phi node. This will insert phi nodes if we cycle in order to
79 // break the cycle and have an operand.
80 bool UniqueIncomingAccess = true;
81 MemoryAccess *SingleAccess = nullptr;
82 for (auto *Pred : predecessors(BB)) {
83 if (MSSA->DT->isReachableFromEntry(Pred)) {
84 auto *IncomingAccess = getPreviousDefFromEnd(Pred, CachedPreviousDef);
85 if (!SingleAccess)
86 SingleAccess = IncomingAccess;
87 else if (IncomingAccess != SingleAccess)
88 UniqueIncomingAccess = false;
89 PhiOps.push_back(IncomingAccess);
90 } else
91 PhiOps.push_back(MSSA->getLiveOnEntryDef());
92 }
93
94 // Now try to simplify the ops to avoid placing a phi.
95 // This may return null if we never created a phi yet, that's okay
96 MemoryPhi *Phi = dyn_cast_or_null<MemoryPhi>(MSSA->getMemoryAccess(BB));
97
98 // See if we can avoid the phi by simplifying it.
99 auto *Result = tryRemoveTrivialPhi(Phi, PhiOps);
100 // If we couldn't simplify, we may have to create a phi
101 if (Result == Phi && UniqueIncomingAccess && SingleAccess) {
102 // A concrete Phi only exists if we created an empty one to break a cycle.
103 if (Phi) {
104 assert(Phi->operands().empty() && "Expected empty Phi");
105 Phi->replaceAllUsesWith(SingleAccess);
106 removeMemoryAccess(Phi);
107 }
108 Result = SingleAccess;
109 } else if (Result == Phi && !(UniqueIncomingAccess && SingleAccess)) {
110 if (!Phi)
111 Phi = MSSA->createMemoryPhi(BB);
112
113 // See if the existing phi operands match what we need.
114 // Unlike normal SSA, we only allow one phi node per block, so we can't just
115 // create a new one.
116 if (Phi->getNumOperands() != 0) {
117 // FIXME: Figure out whether this is dead code and if so remove it.
118 if (!std::equal(Phi->op_begin(), Phi->op_end(), PhiOps.begin())) {
119 // These will have been filled in by the recursive read we did above.
120 llvm::copy(PhiOps, Phi->op_begin());
121 std::copy(pred_begin(BB), pred_end(BB), Phi->block_begin());
122 }
123 } else {
124 unsigned i = 0;
125 for (auto *Pred : predecessors(BB))
126 Phi->addIncoming(&*PhiOps[i++], Pred);
127 InsertedPHIs.push_back(Phi);
128 }
129 Result = Phi;
130 }
131
132 // Set ourselves up for the next variable by resetting visited state.
133 VisitedBlocks.erase(BB);
134 CachedPreviousDef.insert({BB, Result});
135 return Result;
136 }
137 llvm_unreachable("Should have hit one of the three cases above");
138 }
139
140 // This starts at the memory access, and goes backwards in the block to find the
141 // previous definition. If a definition is not found the block of the access,
142 // it continues globally, creating phi nodes to ensure we have a single
143 // definition.
getPreviousDef(MemoryAccess * MA)144 MemoryAccess *MemorySSAUpdater::getPreviousDef(MemoryAccess *MA) {
145 if (auto *LocalResult = getPreviousDefInBlock(MA))
146 return LocalResult;
147 DenseMap<BasicBlock *, TrackingVH<MemoryAccess>> CachedPreviousDef;
148 return getPreviousDefRecursive(MA->getBlock(), CachedPreviousDef);
149 }
150
151 // This starts at the memory access, and goes backwards in the block to the find
152 // the previous definition. If the definition is not found in the block of the
153 // access, it returns nullptr.
getPreviousDefInBlock(MemoryAccess * MA)154 MemoryAccess *MemorySSAUpdater::getPreviousDefInBlock(MemoryAccess *MA) {
155 auto *Defs = MSSA->getWritableBlockDefs(MA->getBlock());
156
157 // It's possible there are no defs, or we got handed the first def to start.
158 if (Defs) {
159 // If this is a def, we can just use the def iterators.
160 if (!isa<MemoryUse>(MA)) {
161 auto Iter = MA->getReverseDefsIterator();
162 ++Iter;
163 if (Iter != Defs->rend())
164 return &*Iter;
165 } else {
166 // Otherwise, have to walk the all access iterator.
167 auto End = MSSA->getWritableBlockAccesses(MA->getBlock())->rend();
168 for (auto &U : make_range(++MA->getReverseIterator(), End))
169 if (!isa<MemoryUse>(U))
170 return cast<MemoryAccess>(&U);
171 // Note that if MA comes before Defs->begin(), we won't hit a def.
172 return nullptr;
173 }
174 }
175 return nullptr;
176 }
177
178 // This starts at the end of block
getPreviousDefFromEnd(BasicBlock * BB,DenseMap<BasicBlock *,TrackingVH<MemoryAccess>> & CachedPreviousDef)179 MemoryAccess *MemorySSAUpdater::getPreviousDefFromEnd(
180 BasicBlock *BB,
181 DenseMap<BasicBlock *, TrackingVH<MemoryAccess>> &CachedPreviousDef) {
182 auto *Defs = MSSA->getWritableBlockDefs(BB);
183
184 if (Defs) {
185 CachedPreviousDef.insert({BB, &*Defs->rbegin()});
186 return &*Defs->rbegin();
187 }
188
189 return getPreviousDefRecursive(BB, CachedPreviousDef);
190 }
191 // Recurse over a set of phi uses to eliminate the trivial ones
recursePhi(MemoryAccess * Phi)192 MemoryAccess *MemorySSAUpdater::recursePhi(MemoryAccess *Phi) {
193 if (!Phi)
194 return nullptr;
195 TrackingVH<MemoryAccess> Res(Phi);
196 SmallVector<TrackingVH<Value>, 8> Uses;
197 std::copy(Phi->user_begin(), Phi->user_end(), std::back_inserter(Uses));
198 for (auto &U : Uses)
199 if (MemoryPhi *UsePhi = dyn_cast<MemoryPhi>(&*U))
200 tryRemoveTrivialPhi(UsePhi);
201 return Res;
202 }
203
204 // Eliminate trivial phis
205 // Phis are trivial if they are defined either by themselves, or all the same
206 // argument.
207 // IE phi(a, a) or b = phi(a, b) or c = phi(a, a, c)
208 // We recursively try to remove them.
tryRemoveTrivialPhi(MemoryPhi * Phi)209 MemoryAccess *MemorySSAUpdater::tryRemoveTrivialPhi(MemoryPhi *Phi) {
210 assert(Phi && "Can only remove concrete Phi.");
211 auto OperRange = Phi->operands();
212 return tryRemoveTrivialPhi(Phi, OperRange);
213 }
214 template <class RangeType>
tryRemoveTrivialPhi(MemoryPhi * Phi,RangeType & Operands)215 MemoryAccess *MemorySSAUpdater::tryRemoveTrivialPhi(MemoryPhi *Phi,
216 RangeType &Operands) {
217 // Bail out on non-opt Phis.
218 if (NonOptPhis.count(Phi))
219 return Phi;
220
221 // Detect equal or self arguments
222 MemoryAccess *Same = nullptr;
223 for (auto &Op : Operands) {
224 // If the same or self, good so far
225 if (Op == Phi || Op == Same)
226 continue;
227 // not the same, return the phi since it's not eliminatable by us
228 if (Same)
229 return Phi;
230 Same = cast<MemoryAccess>(&*Op);
231 }
232 // Never found a non-self reference, the phi is undef
233 if (Same == nullptr)
234 return MSSA->getLiveOnEntryDef();
235 if (Phi) {
236 Phi->replaceAllUsesWith(Same);
237 removeMemoryAccess(Phi);
238 }
239
240 // We should only end up recursing in case we replaced something, in which
241 // case, we may have made other Phis trivial.
242 return recursePhi(Same);
243 }
244
insertUse(MemoryUse * MU,bool RenameUses)245 void MemorySSAUpdater::insertUse(MemoryUse *MU, bool RenameUses) {
246 InsertedPHIs.clear();
247 MU->setDefiningAccess(getPreviousDef(MU));
248
249 // In cases without unreachable blocks, because uses do not create new
250 // may-defs, there are only two cases:
251 // 1. There was a def already below us, and therefore, we should not have
252 // created a phi node because it was already needed for the def.
253 //
254 // 2. There is no def below us, and therefore, there is no extra renaming work
255 // to do.
256
257 // In cases with unreachable blocks, where the unnecessary Phis were
258 // optimized out, adding the Use may re-insert those Phis. Hence, when
259 // inserting Uses outside of the MSSA creation process, and new Phis were
260 // added, rename all uses if we are asked.
261
262 if (!RenameUses && !InsertedPHIs.empty()) {
263 auto *Defs = MSSA->getBlockDefs(MU->getBlock());
264 (void)Defs;
265 assert((!Defs || (++Defs->begin() == Defs->end())) &&
266 "Block may have only a Phi or no defs");
267 }
268
269 if (RenameUses && InsertedPHIs.size()) {
270 SmallPtrSet<BasicBlock *, 16> Visited;
271 BasicBlock *StartBlock = MU->getBlock();
272
273 if (auto *Defs = MSSA->getWritableBlockDefs(StartBlock)) {
274 MemoryAccess *FirstDef = &*Defs->begin();
275 // Convert to incoming value if it's a memorydef. A phi *is* already an
276 // incoming value.
277 if (auto *MD = dyn_cast<MemoryDef>(FirstDef))
278 FirstDef = MD->getDefiningAccess();
279
280 MSSA->renamePass(MU->getBlock(), FirstDef, Visited);
281 }
282 // We just inserted a phi into this block, so the incoming value will
283 // become the phi anyway, so it does not matter what we pass.
284 for (auto &MP : InsertedPHIs)
285 if (MemoryPhi *Phi = cast_or_null<MemoryPhi>(MP))
286 MSSA->renamePass(Phi->getBlock(), nullptr, Visited);
287 }
288 }
289
290 // Set every incoming edge {BB, MP->getBlock()} of MemoryPhi MP to NewDef.
setMemoryPhiValueForBlock(MemoryPhi * MP,const BasicBlock * BB,MemoryAccess * NewDef)291 static void setMemoryPhiValueForBlock(MemoryPhi *MP, const BasicBlock *BB,
292 MemoryAccess *NewDef) {
293 // Replace any operand with us an incoming block with the new defining
294 // access.
295 int i = MP->getBasicBlockIndex(BB);
296 assert(i != -1 && "Should have found the basic block in the phi");
297 // We can't just compare i against getNumOperands since one is signed and the
298 // other not. So use it to index into the block iterator.
299 for (auto BBIter = MP->block_begin() + i; BBIter != MP->block_end();
300 ++BBIter) {
301 if (*BBIter != BB)
302 break;
303 MP->setIncomingValue(i, NewDef);
304 ++i;
305 }
306 }
307
308 // A brief description of the algorithm:
309 // First, we compute what should define the new def, using the SSA
310 // construction algorithm.
311 // Then, we update the defs below us (and any new phi nodes) in the graph to
312 // point to the correct new defs, to ensure we only have one variable, and no
313 // disconnected stores.
insertDef(MemoryDef * MD,bool RenameUses)314 void MemorySSAUpdater::insertDef(MemoryDef *MD, bool RenameUses) {
315 InsertedPHIs.clear();
316
317 // See if we had a local def, and if not, go hunting.
318 MemoryAccess *DefBefore = getPreviousDef(MD);
319 bool DefBeforeSameBlock = false;
320 if (DefBefore->getBlock() == MD->getBlock() &&
321 !(isa<MemoryPhi>(DefBefore) &&
322 llvm::is_contained(InsertedPHIs, DefBefore)))
323 DefBeforeSameBlock = true;
324
325 // There is a def before us, which means we can replace any store/phi uses
326 // of that thing with us, since we are in the way of whatever was there
327 // before.
328 // We now define that def's memorydefs and memoryphis
329 if (DefBeforeSameBlock) {
330 DefBefore->replaceUsesWithIf(MD, [MD](Use &U) {
331 // Leave the MemoryUses alone.
332 // Also make sure we skip ourselves to avoid self references.
333 User *Usr = U.getUser();
334 return !isa<MemoryUse>(Usr) && Usr != MD;
335 // Defs are automatically unoptimized when the user is set to MD below,
336 // because the isOptimized() call will fail to find the same ID.
337 });
338 }
339
340 // and that def is now our defining access.
341 MD->setDefiningAccess(DefBefore);
342
343 SmallVector<WeakVH, 8> FixupList(InsertedPHIs.begin(), InsertedPHIs.end());
344
345 SmallSet<WeakVH, 8> ExistingPhis;
346
347 // Remember the index where we may insert new phis.
348 unsigned NewPhiIndex = InsertedPHIs.size();
349 if (!DefBeforeSameBlock) {
350 // If there was a local def before us, we must have the same effect it
351 // did. Because every may-def is the same, any phis/etc we would create, it
352 // would also have created. If there was no local def before us, we
353 // performed a global update, and have to search all successors and make
354 // sure we update the first def in each of them (following all paths until
355 // we hit the first def along each path). This may also insert phi nodes.
356 // TODO: There are other cases we can skip this work, such as when we have a
357 // single successor, and only used a straight line of single pred blocks
358 // backwards to find the def. To make that work, we'd have to track whether
359 // getDefRecursive only ever used the single predecessor case. These types
360 // of paths also only exist in between CFG simplifications.
361
362 // If this is the first def in the block and this insert is in an arbitrary
363 // place, compute IDF and place phis.
364 SmallPtrSet<BasicBlock *, 2> DefiningBlocks;
365
366 // If this is the last Def in the block, we may need additional Phis.
367 // Compute IDF in all cases, as renaming needs to be done even when MD is
368 // not the last access, because it can introduce a new access past which a
369 // previous access was optimized; that access needs to be reoptimized.
370 DefiningBlocks.insert(MD->getBlock());
371 for (const auto &VH : InsertedPHIs)
372 if (const auto *RealPHI = cast_or_null<MemoryPhi>(VH))
373 DefiningBlocks.insert(RealPHI->getBlock());
374 ForwardIDFCalculator IDFs(*MSSA->DT);
375 SmallVector<BasicBlock *, 32> IDFBlocks;
376 IDFs.setDefiningBlocks(DefiningBlocks);
377 IDFs.calculate(IDFBlocks);
378 SmallVector<AssertingVH<MemoryPhi>, 4> NewInsertedPHIs;
379 for (auto *BBIDF : IDFBlocks) {
380 auto *MPhi = MSSA->getMemoryAccess(BBIDF);
381 if (!MPhi) {
382 MPhi = MSSA->createMemoryPhi(BBIDF);
383 NewInsertedPHIs.push_back(MPhi);
384 } else {
385 ExistingPhis.insert(MPhi);
386 }
387 // Add the phis created into the IDF blocks to NonOptPhis, so they are not
388 // optimized out as trivial by the call to getPreviousDefFromEnd below.
389 // Once they are complete, all these Phis are added to the FixupList, and
390 // removed from NonOptPhis inside fixupDefs(). Existing Phis in IDF may
391 // need fixing as well, and potentially be trivial before this insertion,
392 // hence add all IDF Phis. See PR43044.
393 NonOptPhis.insert(MPhi);
394 }
395 for (auto &MPhi : NewInsertedPHIs) {
396 auto *BBIDF = MPhi->getBlock();
397 for (auto *Pred : predecessors(BBIDF)) {
398 DenseMap<BasicBlock *, TrackingVH<MemoryAccess>> CachedPreviousDef;
399 MPhi->addIncoming(getPreviousDefFromEnd(Pred, CachedPreviousDef), Pred);
400 }
401 }
402
403 // Re-take the index where we're adding the new phis, because the above call
404 // to getPreviousDefFromEnd, may have inserted into InsertedPHIs.
405 NewPhiIndex = InsertedPHIs.size();
406 for (auto &MPhi : NewInsertedPHIs) {
407 InsertedPHIs.push_back(&*MPhi);
408 FixupList.push_back(&*MPhi);
409 }
410
411 FixupList.push_back(MD);
412 }
413
414 // Remember the index where we stopped inserting new phis above, since the
415 // fixupDefs call in the loop below may insert more, that are already minimal.
416 unsigned NewPhiIndexEnd = InsertedPHIs.size();
417
418 while (!FixupList.empty()) {
419 unsigned StartingPHISize = InsertedPHIs.size();
420 fixupDefs(FixupList);
421 FixupList.clear();
422 // Put any new phis on the fixup list, and process them
423 FixupList.append(InsertedPHIs.begin() + StartingPHISize, InsertedPHIs.end());
424 }
425
426 // Optimize potentially non-minimal phis added in this method.
427 unsigned NewPhiSize = NewPhiIndexEnd - NewPhiIndex;
428 if (NewPhiSize)
429 tryRemoveTrivialPhis(ArrayRef<WeakVH>(&InsertedPHIs[NewPhiIndex], NewPhiSize));
430
431 // Now that all fixups are done, rename all uses if we are asked. Skip
432 // renaming for defs in unreachable blocks.
433 BasicBlock *StartBlock = MD->getBlock();
434 if (RenameUses && MSSA->getDomTree().getNode(StartBlock)) {
435 SmallPtrSet<BasicBlock *, 16> Visited;
436 // We are guaranteed there is a def in the block, because we just got it
437 // handed to us in this function.
438 MemoryAccess *FirstDef = &*MSSA->getWritableBlockDefs(StartBlock)->begin();
439 // Convert to incoming value if it's a memorydef. A phi *is* already an
440 // incoming value.
441 if (auto *MD = dyn_cast<MemoryDef>(FirstDef))
442 FirstDef = MD->getDefiningAccess();
443
444 MSSA->renamePass(MD->getBlock(), FirstDef, Visited);
445 // We just inserted a phi into this block, so the incoming value will become
446 // the phi anyway, so it does not matter what we pass.
447 for (auto &MP : InsertedPHIs) {
448 MemoryPhi *Phi = dyn_cast_or_null<MemoryPhi>(MP);
449 if (Phi)
450 MSSA->renamePass(Phi->getBlock(), nullptr, Visited);
451 }
452 // Existing Phi blocks may need renaming too, if an access was previously
453 // optimized and the inserted Defs "covers" the Optimized value.
454 for (auto &MP : ExistingPhis) {
455 MemoryPhi *Phi = dyn_cast_or_null<MemoryPhi>(MP);
456 if (Phi)
457 MSSA->renamePass(Phi->getBlock(), nullptr, Visited);
458 }
459 }
460 }
461
fixupDefs(const SmallVectorImpl<WeakVH> & Vars)462 void MemorySSAUpdater::fixupDefs(const SmallVectorImpl<WeakVH> &Vars) {
463 SmallPtrSet<const BasicBlock *, 8> Seen;
464 SmallVector<const BasicBlock *, 16> Worklist;
465 for (auto &Var : Vars) {
466 MemoryAccess *NewDef = dyn_cast_or_null<MemoryAccess>(Var);
467 if (!NewDef)
468 continue;
469 // First, see if there is a local def after the operand.
470 auto *Defs = MSSA->getWritableBlockDefs(NewDef->getBlock());
471 auto DefIter = NewDef->getDefsIterator();
472
473 // The temporary Phi is being fixed, unmark it for not to optimize.
474 if (MemoryPhi *Phi = dyn_cast<MemoryPhi>(NewDef))
475 NonOptPhis.erase(Phi);
476
477 // If there is a local def after us, we only have to rename that.
478 if (++DefIter != Defs->end()) {
479 cast<MemoryDef>(DefIter)->setDefiningAccess(NewDef);
480 continue;
481 }
482
483 // Otherwise, we need to search down through the CFG.
484 // For each of our successors, handle it directly if their is a phi, or
485 // place on the fixup worklist.
486 for (const auto *S : successors(NewDef->getBlock())) {
487 if (auto *MP = MSSA->getMemoryAccess(S))
488 setMemoryPhiValueForBlock(MP, NewDef->getBlock(), NewDef);
489 else
490 Worklist.push_back(S);
491 }
492
493 while (!Worklist.empty()) {
494 const BasicBlock *FixupBlock = Worklist.back();
495 Worklist.pop_back();
496
497 // Get the first def in the block that isn't a phi node.
498 if (auto *Defs = MSSA->getWritableBlockDefs(FixupBlock)) {
499 auto *FirstDef = &*Defs->begin();
500 // The loop above and below should have taken care of phi nodes
501 assert(!isa<MemoryPhi>(FirstDef) &&
502 "Should have already handled phi nodes!");
503 // We are now this def's defining access, make sure we actually dominate
504 // it
505 assert(MSSA->dominates(NewDef, FirstDef) &&
506 "Should have dominated the new access");
507
508 // This may insert new phi nodes, because we are not guaranteed the
509 // block we are processing has a single pred, and depending where the
510 // store was inserted, it may require phi nodes below it.
511 cast<MemoryDef>(FirstDef)->setDefiningAccess(getPreviousDef(FirstDef));
512 return;
513 }
514 // We didn't find a def, so we must continue.
515 for (const auto *S : successors(FixupBlock)) {
516 // If there is a phi node, handle it.
517 // Otherwise, put the block on the worklist
518 if (auto *MP = MSSA->getMemoryAccess(S))
519 setMemoryPhiValueForBlock(MP, FixupBlock, NewDef);
520 else {
521 // If we cycle, we should have ended up at a phi node that we already
522 // processed. FIXME: Double check this
523 if (!Seen.insert(S).second)
524 continue;
525 Worklist.push_back(S);
526 }
527 }
528 }
529 }
530 }
531
removeEdge(BasicBlock * From,BasicBlock * To)532 void MemorySSAUpdater::removeEdge(BasicBlock *From, BasicBlock *To) {
533 if (MemoryPhi *MPhi = MSSA->getMemoryAccess(To)) {
534 MPhi->unorderedDeleteIncomingBlock(From);
535 tryRemoveTrivialPhi(MPhi);
536 }
537 }
538
removeDuplicatePhiEdgesBetween(const BasicBlock * From,const BasicBlock * To)539 void MemorySSAUpdater::removeDuplicatePhiEdgesBetween(const BasicBlock *From,
540 const BasicBlock *To) {
541 if (MemoryPhi *MPhi = MSSA->getMemoryAccess(To)) {
542 bool Found = false;
543 MPhi->unorderedDeleteIncomingIf([&](const MemoryAccess *, BasicBlock *B) {
544 if (From != B)
545 return false;
546 if (Found)
547 return true;
548 Found = true;
549 return false;
550 });
551 tryRemoveTrivialPhi(MPhi);
552 }
553 }
554
555 /// If all arguments of a MemoryPHI are defined by the same incoming
556 /// argument, return that argument.
onlySingleValue(MemoryPhi * MP)557 static MemoryAccess *onlySingleValue(MemoryPhi *MP) {
558 MemoryAccess *MA = nullptr;
559
560 for (auto &Arg : MP->operands()) {
561 if (!MA)
562 MA = cast<MemoryAccess>(Arg);
563 else if (MA != Arg)
564 return nullptr;
565 }
566 return MA;
567 }
568
getNewDefiningAccessForClone(MemoryAccess * MA,const ValueToValueMapTy & VMap,PhiToDefMap & MPhiMap,bool CloneWasSimplified,MemorySSA * MSSA)569 static MemoryAccess *getNewDefiningAccessForClone(MemoryAccess *MA,
570 const ValueToValueMapTy &VMap,
571 PhiToDefMap &MPhiMap,
572 bool CloneWasSimplified,
573 MemorySSA *MSSA) {
574 MemoryAccess *InsnDefining = MA;
575 if (MemoryDef *DefMUD = dyn_cast<MemoryDef>(InsnDefining)) {
576 if (!MSSA->isLiveOnEntryDef(DefMUD)) {
577 Instruction *DefMUDI = DefMUD->getMemoryInst();
578 assert(DefMUDI && "Found MemoryUseOrDef with no Instruction.");
579 if (Instruction *NewDefMUDI =
580 cast_or_null<Instruction>(VMap.lookup(DefMUDI))) {
581 InsnDefining = MSSA->getMemoryAccess(NewDefMUDI);
582 if (!CloneWasSimplified)
583 assert(InsnDefining && "Defining instruction cannot be nullptr.");
584 else if (!InsnDefining || isa<MemoryUse>(InsnDefining)) {
585 // The clone was simplified, it's no longer a MemoryDef, look up.
586 auto DefIt = DefMUD->getDefsIterator();
587 // Since simplified clones only occur in single block cloning, a
588 // previous definition must exist, otherwise NewDefMUDI would not
589 // have been found in VMap.
590 assert(DefIt != MSSA->getBlockDefs(DefMUD->getBlock())->begin() &&
591 "Previous def must exist");
592 InsnDefining = getNewDefiningAccessForClone(
593 &*(--DefIt), VMap, MPhiMap, CloneWasSimplified, MSSA);
594 }
595 }
596 }
597 } else {
598 MemoryPhi *DefPhi = cast<MemoryPhi>(InsnDefining);
599 if (MemoryAccess *NewDefPhi = MPhiMap.lookup(DefPhi))
600 InsnDefining = NewDefPhi;
601 }
602 assert(InsnDefining && "Defining instruction cannot be nullptr.");
603 return InsnDefining;
604 }
605
cloneUsesAndDefs(BasicBlock * BB,BasicBlock * NewBB,const ValueToValueMapTy & VMap,PhiToDefMap & MPhiMap,bool CloneWasSimplified)606 void MemorySSAUpdater::cloneUsesAndDefs(BasicBlock *BB, BasicBlock *NewBB,
607 const ValueToValueMapTy &VMap,
608 PhiToDefMap &MPhiMap,
609 bool CloneWasSimplified) {
610 const MemorySSA::AccessList *Acc = MSSA->getBlockAccesses(BB);
611 if (!Acc)
612 return;
613 for (const MemoryAccess &MA : *Acc) {
614 if (const MemoryUseOrDef *MUD = dyn_cast<MemoryUseOrDef>(&MA)) {
615 Instruction *Insn = MUD->getMemoryInst();
616 // Entry does not exist if the clone of the block did not clone all
617 // instructions. This occurs in LoopRotate when cloning instructions
618 // from the old header to the old preheader. The cloned instruction may
619 // also be a simplified Value, not an Instruction (see LoopRotate).
620 // Also in LoopRotate, even when it's an instruction, due to it being
621 // simplified, it may be a Use rather than a Def, so we cannot use MUD as
622 // template. Calls coming from updateForClonedBlockIntoPred, ensure this.
623 if (Instruction *NewInsn =
624 dyn_cast_or_null<Instruction>(VMap.lookup(Insn))) {
625 MemoryAccess *NewUseOrDef = MSSA->createDefinedAccess(
626 NewInsn,
627 getNewDefiningAccessForClone(MUD->getDefiningAccess(), VMap,
628 MPhiMap, CloneWasSimplified, MSSA),
629 /*Template=*/CloneWasSimplified ? nullptr : MUD,
630 /*CreationMustSucceed=*/CloneWasSimplified ? false : true);
631 if (NewUseOrDef)
632 MSSA->insertIntoListsForBlock(NewUseOrDef, NewBB, MemorySSA::End);
633 }
634 }
635 }
636 }
637
updatePhisWhenInsertingUniqueBackedgeBlock(BasicBlock * Header,BasicBlock * Preheader,BasicBlock * BEBlock)638 void MemorySSAUpdater::updatePhisWhenInsertingUniqueBackedgeBlock(
639 BasicBlock *Header, BasicBlock *Preheader, BasicBlock *BEBlock) {
640 auto *MPhi = MSSA->getMemoryAccess(Header);
641 if (!MPhi)
642 return;
643
644 // Create phi node in the backedge block and populate it with the same
645 // incoming values as MPhi. Skip incoming values coming from Preheader.
646 auto *NewMPhi = MSSA->createMemoryPhi(BEBlock);
647 bool HasUniqueIncomingValue = true;
648 MemoryAccess *UniqueValue = nullptr;
649 for (unsigned I = 0, E = MPhi->getNumIncomingValues(); I != E; ++I) {
650 BasicBlock *IBB = MPhi->getIncomingBlock(I);
651 MemoryAccess *IV = MPhi->getIncomingValue(I);
652 if (IBB != Preheader) {
653 NewMPhi->addIncoming(IV, IBB);
654 if (HasUniqueIncomingValue) {
655 if (!UniqueValue)
656 UniqueValue = IV;
657 else if (UniqueValue != IV)
658 HasUniqueIncomingValue = false;
659 }
660 }
661 }
662
663 // Update incoming edges into MPhi. Remove all but the incoming edge from
664 // Preheader. Add an edge from NewMPhi
665 auto *AccFromPreheader = MPhi->getIncomingValueForBlock(Preheader);
666 MPhi->setIncomingValue(0, AccFromPreheader);
667 MPhi->setIncomingBlock(0, Preheader);
668 for (unsigned I = MPhi->getNumIncomingValues() - 1; I >= 1; --I)
669 MPhi->unorderedDeleteIncoming(I);
670 MPhi->addIncoming(NewMPhi, BEBlock);
671
672 // If NewMPhi is a trivial phi, remove it. Its use in the header MPhi will be
673 // replaced with the unique value.
674 tryRemoveTrivialPhi(NewMPhi);
675 }
676
updateForClonedLoop(const LoopBlocksRPO & LoopBlocks,ArrayRef<BasicBlock * > ExitBlocks,const ValueToValueMapTy & VMap,bool IgnoreIncomingWithNoClones)677 void MemorySSAUpdater::updateForClonedLoop(const LoopBlocksRPO &LoopBlocks,
678 ArrayRef<BasicBlock *> ExitBlocks,
679 const ValueToValueMapTy &VMap,
680 bool IgnoreIncomingWithNoClones) {
681 PhiToDefMap MPhiMap;
682
683 auto FixPhiIncomingValues = [&](MemoryPhi *Phi, MemoryPhi *NewPhi) {
684 assert(Phi && NewPhi && "Invalid Phi nodes.");
685 BasicBlock *NewPhiBB = NewPhi->getBlock();
686 SmallPtrSet<BasicBlock *, 4> NewPhiBBPreds(pred_begin(NewPhiBB),
687 pred_end(NewPhiBB));
688 for (unsigned It = 0, E = Phi->getNumIncomingValues(); It < E; ++It) {
689 MemoryAccess *IncomingAccess = Phi->getIncomingValue(It);
690 BasicBlock *IncBB = Phi->getIncomingBlock(It);
691
692 if (BasicBlock *NewIncBB = cast_or_null<BasicBlock>(VMap.lookup(IncBB)))
693 IncBB = NewIncBB;
694 else if (IgnoreIncomingWithNoClones)
695 continue;
696
697 // Now we have IncBB, and will need to add incoming from it to NewPhi.
698
699 // If IncBB is not a predecessor of NewPhiBB, then do not add it.
700 // NewPhiBB was cloned without that edge.
701 if (!NewPhiBBPreds.count(IncBB))
702 continue;
703
704 // Determine incoming value and add it as incoming from IncBB.
705 if (MemoryUseOrDef *IncMUD = dyn_cast<MemoryUseOrDef>(IncomingAccess)) {
706 if (!MSSA->isLiveOnEntryDef(IncMUD)) {
707 Instruction *IncI = IncMUD->getMemoryInst();
708 assert(IncI && "Found MemoryUseOrDef with no Instruction.");
709 if (Instruction *NewIncI =
710 cast_or_null<Instruction>(VMap.lookup(IncI))) {
711 IncMUD = MSSA->getMemoryAccess(NewIncI);
712 assert(IncMUD &&
713 "MemoryUseOrDef cannot be null, all preds processed.");
714 }
715 }
716 NewPhi->addIncoming(IncMUD, IncBB);
717 } else {
718 MemoryPhi *IncPhi = cast<MemoryPhi>(IncomingAccess);
719 if (MemoryAccess *NewDefPhi = MPhiMap.lookup(IncPhi))
720 NewPhi->addIncoming(NewDefPhi, IncBB);
721 else
722 NewPhi->addIncoming(IncPhi, IncBB);
723 }
724 }
725 if (auto *SingleAccess = onlySingleValue(NewPhi)) {
726 MPhiMap[Phi] = SingleAccess;
727 removeMemoryAccess(NewPhi);
728 }
729 };
730
731 auto ProcessBlock = [&](BasicBlock *BB) {
732 BasicBlock *NewBlock = cast_or_null<BasicBlock>(VMap.lookup(BB));
733 if (!NewBlock)
734 return;
735
736 assert(!MSSA->getWritableBlockAccesses(NewBlock) &&
737 "Cloned block should have no accesses");
738
739 // Add MemoryPhi.
740 if (MemoryPhi *MPhi = MSSA->getMemoryAccess(BB)) {
741 MemoryPhi *NewPhi = MSSA->createMemoryPhi(NewBlock);
742 MPhiMap[MPhi] = NewPhi;
743 }
744 // Update Uses and Defs.
745 cloneUsesAndDefs(BB, NewBlock, VMap, MPhiMap);
746 };
747
748 for (auto BB : llvm::concat<BasicBlock *const>(LoopBlocks, ExitBlocks))
749 ProcessBlock(BB);
750
751 for (auto BB : llvm::concat<BasicBlock *const>(LoopBlocks, ExitBlocks))
752 if (MemoryPhi *MPhi = MSSA->getMemoryAccess(BB))
753 if (MemoryAccess *NewPhi = MPhiMap.lookup(MPhi))
754 FixPhiIncomingValues(MPhi, cast<MemoryPhi>(NewPhi));
755 }
756
updateForClonedBlockIntoPred(BasicBlock * BB,BasicBlock * P1,const ValueToValueMapTy & VM)757 void MemorySSAUpdater::updateForClonedBlockIntoPred(
758 BasicBlock *BB, BasicBlock *P1, const ValueToValueMapTy &VM) {
759 // All defs/phis from outside BB that are used in BB, are valid uses in P1.
760 // Since those defs/phis must have dominated BB, and also dominate P1.
761 // Defs from BB being used in BB will be replaced with the cloned defs from
762 // VM. The uses of BB's Phi (if it exists) in BB will be replaced by the
763 // incoming def into the Phi from P1.
764 // Instructions cloned into the predecessor are in practice sometimes
765 // simplified, so disable the use of the template, and create an access from
766 // scratch.
767 PhiToDefMap MPhiMap;
768 if (MemoryPhi *MPhi = MSSA->getMemoryAccess(BB))
769 MPhiMap[MPhi] = MPhi->getIncomingValueForBlock(P1);
770 cloneUsesAndDefs(BB, P1, VM, MPhiMap, /*CloneWasSimplified=*/true);
771 }
772
773 template <typename Iter>
privateUpdateExitBlocksForClonedLoop(ArrayRef<BasicBlock * > ExitBlocks,Iter ValuesBegin,Iter ValuesEnd,DominatorTree & DT)774 void MemorySSAUpdater::privateUpdateExitBlocksForClonedLoop(
775 ArrayRef<BasicBlock *> ExitBlocks, Iter ValuesBegin, Iter ValuesEnd,
776 DominatorTree &DT) {
777 SmallVector<CFGUpdate, 4> Updates;
778 // Update/insert phis in all successors of exit blocks.
779 for (auto *Exit : ExitBlocks)
780 for (const ValueToValueMapTy *VMap : make_range(ValuesBegin, ValuesEnd))
781 if (BasicBlock *NewExit = cast_or_null<BasicBlock>(VMap->lookup(Exit))) {
782 BasicBlock *ExitSucc = NewExit->getTerminator()->getSuccessor(0);
783 Updates.push_back({DT.Insert, NewExit, ExitSucc});
784 }
785 applyInsertUpdates(Updates, DT);
786 }
787
updateExitBlocksForClonedLoop(ArrayRef<BasicBlock * > ExitBlocks,const ValueToValueMapTy & VMap,DominatorTree & DT)788 void MemorySSAUpdater::updateExitBlocksForClonedLoop(
789 ArrayRef<BasicBlock *> ExitBlocks, const ValueToValueMapTy &VMap,
790 DominatorTree &DT) {
791 const ValueToValueMapTy *const Arr[] = {&VMap};
792 privateUpdateExitBlocksForClonedLoop(ExitBlocks, std::begin(Arr),
793 std::end(Arr), DT);
794 }
795
updateExitBlocksForClonedLoop(ArrayRef<BasicBlock * > ExitBlocks,ArrayRef<std::unique_ptr<ValueToValueMapTy>> VMaps,DominatorTree & DT)796 void MemorySSAUpdater::updateExitBlocksForClonedLoop(
797 ArrayRef<BasicBlock *> ExitBlocks,
798 ArrayRef<std::unique_ptr<ValueToValueMapTy>> VMaps, DominatorTree &DT) {
799 auto GetPtr = [&](const std::unique_ptr<ValueToValueMapTy> &I) {
800 return I.get();
801 };
802 using MappedIteratorType =
803 mapped_iterator<const std::unique_ptr<ValueToValueMapTy> *,
804 decltype(GetPtr)>;
805 auto MapBegin = MappedIteratorType(VMaps.begin(), GetPtr);
806 auto MapEnd = MappedIteratorType(VMaps.end(), GetPtr);
807 privateUpdateExitBlocksForClonedLoop(ExitBlocks, MapBegin, MapEnd, DT);
808 }
809
applyUpdates(ArrayRef<CFGUpdate> Updates,DominatorTree & DT,bool UpdateDT)810 void MemorySSAUpdater::applyUpdates(ArrayRef<CFGUpdate> Updates,
811 DominatorTree &DT, bool UpdateDT) {
812 SmallVector<CFGUpdate, 4> DeleteUpdates;
813 SmallVector<CFGUpdate, 4> RevDeleteUpdates;
814 SmallVector<CFGUpdate, 4> InsertUpdates;
815 for (auto &Update : Updates) {
816 if (Update.getKind() == DT.Insert)
817 InsertUpdates.push_back({DT.Insert, Update.getFrom(), Update.getTo()});
818 else {
819 DeleteUpdates.push_back({DT.Delete, Update.getFrom(), Update.getTo()});
820 RevDeleteUpdates.push_back({DT.Insert, Update.getFrom(), Update.getTo()});
821 }
822 }
823
824 if (!DeleteUpdates.empty()) {
825 if (!UpdateDT) {
826 SmallVector<CFGUpdate, 0> Empty;
827 // Deletes are reversed applied, because this CFGView is pretending the
828 // deletes did not happen yet, hence the edges still exist.
829 DT.applyUpdates(Empty, RevDeleteUpdates);
830 } else {
831 // Apply all updates, with the RevDeleteUpdates as PostCFGView.
832 DT.applyUpdates(Updates, RevDeleteUpdates);
833 }
834
835 // Note: the MSSA update below doesn't distinguish between a GD with
836 // (RevDelete,false) and (Delete, true), but this matters for the DT
837 // updates above; for "children" purposes they are equivalent; but the
838 // updates themselves convey the desired update, used inside DT only.
839 GraphDiff<BasicBlock *> GD(RevDeleteUpdates);
840 applyInsertUpdates(InsertUpdates, DT, &GD);
841 // Update DT to redelete edges; this matches the real CFG so we can perform
842 // the standard update without a postview of the CFG.
843 DT.applyUpdates(DeleteUpdates);
844 } else {
845 if (UpdateDT)
846 DT.applyUpdates(Updates);
847 GraphDiff<BasicBlock *> GD;
848 applyInsertUpdates(InsertUpdates, DT, &GD);
849 }
850
851 // Update for deleted edges
852 for (auto &Update : DeleteUpdates)
853 removeEdge(Update.getFrom(), Update.getTo());
854 }
855
applyInsertUpdates(ArrayRef<CFGUpdate> Updates,DominatorTree & DT)856 void MemorySSAUpdater::applyInsertUpdates(ArrayRef<CFGUpdate> Updates,
857 DominatorTree &DT) {
858 GraphDiff<BasicBlock *> GD;
859 applyInsertUpdates(Updates, DT, &GD);
860 }
861
applyInsertUpdates(ArrayRef<CFGUpdate> Updates,DominatorTree & DT,const GraphDiff<BasicBlock * > * GD)862 void MemorySSAUpdater::applyInsertUpdates(ArrayRef<CFGUpdate> Updates,
863 DominatorTree &DT,
864 const GraphDiff<BasicBlock *> *GD) {
865 // Get recursive last Def, assuming well formed MSSA and updated DT.
866 auto GetLastDef = [&](BasicBlock *BB) -> MemoryAccess * {
867 while (true) {
868 MemorySSA::DefsList *Defs = MSSA->getWritableBlockDefs(BB);
869 // Return last Def or Phi in BB, if it exists.
870 if (Defs)
871 return &*(--Defs->end());
872
873 // Check number of predecessors, we only care if there's more than one.
874 unsigned Count = 0;
875 BasicBlock *Pred = nullptr;
876 for (auto *Pi : GD->template getChildren</*InverseEdge=*/true>(BB)) {
877 Pred = Pi;
878 Count++;
879 if (Count == 2)
880 break;
881 }
882
883 // If BB has multiple predecessors, get last definition from IDom.
884 if (Count != 1) {
885 // [SimpleLoopUnswitch] If BB is a dead block, about to be deleted, its
886 // DT is invalidated. Return LoE as its last def. This will be added to
887 // MemoryPhi node, and later deleted when the block is deleted.
888 if (!DT.getNode(BB))
889 return MSSA->getLiveOnEntryDef();
890 if (auto *IDom = DT.getNode(BB)->getIDom())
891 if (IDom->getBlock() != BB) {
892 BB = IDom->getBlock();
893 continue;
894 }
895 return MSSA->getLiveOnEntryDef();
896 } else {
897 // Single predecessor, BB cannot be dead. GetLastDef of Pred.
898 assert(Count == 1 && Pred && "Single predecessor expected.");
899 // BB can be unreachable though, return LoE if that is the case.
900 if (!DT.getNode(BB))
901 return MSSA->getLiveOnEntryDef();
902 BB = Pred;
903 }
904 };
905 llvm_unreachable("Unable to get last definition.");
906 };
907
908 // Get nearest IDom given a set of blocks.
909 // TODO: this can be optimized by starting the search at the node with the
910 // lowest level (highest in the tree).
911 auto FindNearestCommonDominator =
912 [&](const SmallSetVector<BasicBlock *, 2> &BBSet) -> BasicBlock * {
913 BasicBlock *PrevIDom = *BBSet.begin();
914 for (auto *BB : BBSet)
915 PrevIDom = DT.findNearestCommonDominator(PrevIDom, BB);
916 return PrevIDom;
917 };
918
919 // Get all blocks that dominate PrevIDom, stop when reaching CurrIDom. Do not
920 // include CurrIDom.
921 auto GetNoLongerDomBlocks =
922 [&](BasicBlock *PrevIDom, BasicBlock *CurrIDom,
923 SmallVectorImpl<BasicBlock *> &BlocksPrevDom) {
924 if (PrevIDom == CurrIDom)
925 return;
926 BlocksPrevDom.push_back(PrevIDom);
927 BasicBlock *NextIDom = PrevIDom;
928 while (BasicBlock *UpIDom =
929 DT.getNode(NextIDom)->getIDom()->getBlock()) {
930 if (UpIDom == CurrIDom)
931 break;
932 BlocksPrevDom.push_back(UpIDom);
933 NextIDom = UpIDom;
934 }
935 };
936
937 // Map a BB to its predecessors: added + previously existing. To get a
938 // deterministic order, store predecessors as SetVectors. The order in each
939 // will be defined by the order in Updates (fixed) and the order given by
940 // children<> (also fixed). Since we further iterate over these ordered sets,
941 // we lose the information of multiple edges possibly existing between two
942 // blocks, so we'll keep and EdgeCount map for that.
943 // An alternate implementation could keep unordered set for the predecessors,
944 // traverse either Updates or children<> each time to get the deterministic
945 // order, and drop the usage of EdgeCount. This alternate approach would still
946 // require querying the maps for each predecessor, and children<> call has
947 // additional computation inside for creating the snapshot-graph predecessors.
948 // As such, we favor using a little additional storage and less compute time.
949 // This decision can be revisited if we find the alternative more favorable.
950
951 struct PredInfo {
952 SmallSetVector<BasicBlock *, 2> Added;
953 SmallSetVector<BasicBlock *, 2> Prev;
954 };
955 SmallDenseMap<BasicBlock *, PredInfo> PredMap;
956
957 for (auto &Edge : Updates) {
958 BasicBlock *BB = Edge.getTo();
959 auto &AddedBlockSet = PredMap[BB].Added;
960 AddedBlockSet.insert(Edge.getFrom());
961 }
962
963 // Store all existing predecessor for each BB, at least one must exist.
964 SmallDenseMap<std::pair<BasicBlock *, BasicBlock *>, int> EdgeCountMap;
965 SmallPtrSet<BasicBlock *, 2> NewBlocks;
966 for (auto &BBPredPair : PredMap) {
967 auto *BB = BBPredPair.first;
968 const auto &AddedBlockSet = BBPredPair.second.Added;
969 auto &PrevBlockSet = BBPredPair.second.Prev;
970 for (auto *Pi : GD->template getChildren</*InverseEdge=*/true>(BB)) {
971 if (!AddedBlockSet.count(Pi))
972 PrevBlockSet.insert(Pi);
973 EdgeCountMap[{Pi, BB}]++;
974 }
975
976 if (PrevBlockSet.empty()) {
977 assert(pred_size(BB) == AddedBlockSet.size() && "Duplicate edges added.");
978 LLVM_DEBUG(
979 dbgs()
980 << "Adding a predecessor to a block with no predecessors. "
981 "This must be an edge added to a new, likely cloned, block. "
982 "Its memory accesses must be already correct, assuming completed "
983 "via the updateExitBlocksForClonedLoop API. "
984 "Assert a single such edge is added so no phi addition or "
985 "additional processing is required.\n");
986 assert(AddedBlockSet.size() == 1 &&
987 "Can only handle adding one predecessor to a new block.");
988 // Need to remove new blocks from PredMap. Remove below to not invalidate
989 // iterator here.
990 NewBlocks.insert(BB);
991 }
992 }
993 // Nothing to process for new/cloned blocks.
994 for (auto *BB : NewBlocks)
995 PredMap.erase(BB);
996
997 SmallVector<BasicBlock *, 16> BlocksWithDefsToReplace;
998 SmallVector<WeakVH, 8> InsertedPhis;
999
1000 // First create MemoryPhis in all blocks that don't have one. Create in the
1001 // order found in Updates, not in PredMap, to get deterministic numbering.
1002 for (auto &Edge : Updates) {
1003 BasicBlock *BB = Edge.getTo();
1004 if (PredMap.count(BB) && !MSSA->getMemoryAccess(BB))
1005 InsertedPhis.push_back(MSSA->createMemoryPhi(BB));
1006 }
1007
1008 // Now we'll fill in the MemoryPhis with the right incoming values.
1009 for (auto &BBPredPair : PredMap) {
1010 auto *BB = BBPredPair.first;
1011 const auto &PrevBlockSet = BBPredPair.second.Prev;
1012 const auto &AddedBlockSet = BBPredPair.second.Added;
1013 assert(!PrevBlockSet.empty() &&
1014 "At least one previous predecessor must exist.");
1015
1016 // TODO: if this becomes a bottleneck, we can save on GetLastDef calls by
1017 // keeping this map before the loop. We can reuse already populated entries
1018 // if an edge is added from the same predecessor to two different blocks,
1019 // and this does happen in rotate. Note that the map needs to be updated
1020 // when deleting non-necessary phis below, if the phi is in the map by
1021 // replacing the value with DefP1.
1022 SmallDenseMap<BasicBlock *, MemoryAccess *> LastDefAddedPred;
1023 for (auto *AddedPred : AddedBlockSet) {
1024 auto *DefPn = GetLastDef(AddedPred);
1025 assert(DefPn != nullptr && "Unable to find last definition.");
1026 LastDefAddedPred[AddedPred] = DefPn;
1027 }
1028
1029 MemoryPhi *NewPhi = MSSA->getMemoryAccess(BB);
1030 // If Phi is not empty, add an incoming edge from each added pred. Must
1031 // still compute blocks with defs to replace for this block below.
1032 if (NewPhi->getNumOperands()) {
1033 for (auto *Pred : AddedBlockSet) {
1034 auto *LastDefForPred = LastDefAddedPred[Pred];
1035 for (int I = 0, E = EdgeCountMap[{Pred, BB}]; I < E; ++I)
1036 NewPhi->addIncoming(LastDefForPred, Pred);
1037 }
1038 } else {
1039 // Pick any existing predecessor and get its definition. All other
1040 // existing predecessors should have the same one, since no phi existed.
1041 auto *P1 = *PrevBlockSet.begin();
1042 MemoryAccess *DefP1 = GetLastDef(P1);
1043
1044 // Check DefP1 against all Defs in LastDefPredPair. If all the same,
1045 // nothing to add.
1046 bool InsertPhi = false;
1047 for (auto LastDefPredPair : LastDefAddedPred)
1048 if (DefP1 != LastDefPredPair.second) {
1049 InsertPhi = true;
1050 break;
1051 }
1052 if (!InsertPhi) {
1053 // Since NewPhi may be used in other newly added Phis, replace all uses
1054 // of NewPhi with the definition coming from all predecessors (DefP1),
1055 // before deleting it.
1056 NewPhi->replaceAllUsesWith(DefP1);
1057 removeMemoryAccess(NewPhi);
1058 continue;
1059 }
1060
1061 // Update Phi with new values for new predecessors and old value for all
1062 // other predecessors. Since AddedBlockSet and PrevBlockSet are ordered
1063 // sets, the order of entries in NewPhi is deterministic.
1064 for (auto *Pred : AddedBlockSet) {
1065 auto *LastDefForPred = LastDefAddedPred[Pred];
1066 for (int I = 0, E = EdgeCountMap[{Pred, BB}]; I < E; ++I)
1067 NewPhi->addIncoming(LastDefForPred, Pred);
1068 }
1069 for (auto *Pred : PrevBlockSet)
1070 for (int I = 0, E = EdgeCountMap[{Pred, BB}]; I < E; ++I)
1071 NewPhi->addIncoming(DefP1, Pred);
1072 }
1073
1074 // Get all blocks that used to dominate BB and no longer do after adding
1075 // AddedBlockSet, where PrevBlockSet are the previously known predecessors.
1076 assert(DT.getNode(BB)->getIDom() && "BB does not have valid idom");
1077 BasicBlock *PrevIDom = FindNearestCommonDominator(PrevBlockSet);
1078 assert(PrevIDom && "Previous IDom should exists");
1079 BasicBlock *NewIDom = DT.getNode(BB)->getIDom()->getBlock();
1080 assert(NewIDom && "BB should have a new valid idom");
1081 assert(DT.dominates(NewIDom, PrevIDom) &&
1082 "New idom should dominate old idom");
1083 GetNoLongerDomBlocks(PrevIDom, NewIDom, BlocksWithDefsToReplace);
1084 }
1085
1086 tryRemoveTrivialPhis(InsertedPhis);
1087 // Create the set of blocks that now have a definition. We'll use this to
1088 // compute IDF and add Phis there next.
1089 SmallVector<BasicBlock *, 8> BlocksToProcess;
1090 for (auto &VH : InsertedPhis)
1091 if (auto *MPhi = cast_or_null<MemoryPhi>(VH))
1092 BlocksToProcess.push_back(MPhi->getBlock());
1093
1094 // Compute IDF and add Phis in all IDF blocks that do not have one.
1095 SmallVector<BasicBlock *, 32> IDFBlocks;
1096 if (!BlocksToProcess.empty()) {
1097 ForwardIDFCalculator IDFs(DT, GD);
1098 SmallPtrSet<BasicBlock *, 16> DefiningBlocks(BlocksToProcess.begin(),
1099 BlocksToProcess.end());
1100 IDFs.setDefiningBlocks(DefiningBlocks);
1101 IDFs.calculate(IDFBlocks);
1102
1103 SmallSetVector<MemoryPhi *, 4> PhisToFill;
1104 // First create all needed Phis.
1105 for (auto *BBIDF : IDFBlocks)
1106 if (!MSSA->getMemoryAccess(BBIDF)) {
1107 auto *IDFPhi = MSSA->createMemoryPhi(BBIDF);
1108 InsertedPhis.push_back(IDFPhi);
1109 PhisToFill.insert(IDFPhi);
1110 }
1111 // Then update or insert their correct incoming values.
1112 for (auto *BBIDF : IDFBlocks) {
1113 auto *IDFPhi = MSSA->getMemoryAccess(BBIDF);
1114 assert(IDFPhi && "Phi must exist");
1115 if (!PhisToFill.count(IDFPhi)) {
1116 // Update existing Phi.
1117 // FIXME: some updates may be redundant, try to optimize and skip some.
1118 for (unsigned I = 0, E = IDFPhi->getNumIncomingValues(); I < E; ++I)
1119 IDFPhi->setIncomingValue(I, GetLastDef(IDFPhi->getIncomingBlock(I)));
1120 } else {
1121 for (auto *Pi : GD->template getChildren</*InverseEdge=*/true>(BBIDF))
1122 IDFPhi->addIncoming(GetLastDef(Pi), Pi);
1123 }
1124 }
1125 }
1126
1127 // Now for all defs in BlocksWithDefsToReplace, if there are uses they no
1128 // longer dominate, replace those with the closest dominating def.
1129 // This will also update optimized accesses, as they're also uses.
1130 for (auto *BlockWithDefsToReplace : BlocksWithDefsToReplace) {
1131 if (auto DefsList = MSSA->getWritableBlockDefs(BlockWithDefsToReplace)) {
1132 for (auto &DefToReplaceUses : *DefsList) {
1133 BasicBlock *DominatingBlock = DefToReplaceUses.getBlock();
1134 Value::use_iterator UI = DefToReplaceUses.use_begin(),
1135 E = DefToReplaceUses.use_end();
1136 for (; UI != E;) {
1137 Use &U = *UI;
1138 ++UI;
1139 MemoryAccess *Usr = cast<MemoryAccess>(U.getUser());
1140 if (MemoryPhi *UsrPhi = dyn_cast<MemoryPhi>(Usr)) {
1141 BasicBlock *DominatedBlock = UsrPhi->getIncomingBlock(U);
1142 if (!DT.dominates(DominatingBlock, DominatedBlock))
1143 U.set(GetLastDef(DominatedBlock));
1144 } else {
1145 BasicBlock *DominatedBlock = Usr->getBlock();
1146 if (!DT.dominates(DominatingBlock, DominatedBlock)) {
1147 if (auto *DomBlPhi = MSSA->getMemoryAccess(DominatedBlock))
1148 U.set(DomBlPhi);
1149 else {
1150 auto *IDom = DT.getNode(DominatedBlock)->getIDom();
1151 assert(IDom && "Block must have a valid IDom.");
1152 U.set(GetLastDef(IDom->getBlock()));
1153 }
1154 cast<MemoryUseOrDef>(Usr)->resetOptimized();
1155 }
1156 }
1157 }
1158 }
1159 }
1160 }
1161 tryRemoveTrivialPhis(InsertedPhis);
1162 }
1163
1164 // Move What before Where in the MemorySSA IR.
1165 template <class WhereType>
moveTo(MemoryUseOrDef * What,BasicBlock * BB,WhereType Where)1166 void MemorySSAUpdater::moveTo(MemoryUseOrDef *What, BasicBlock *BB,
1167 WhereType Where) {
1168 // Mark MemoryPhi users of What not to be optimized.
1169 for (auto *U : What->users())
1170 if (MemoryPhi *PhiUser = dyn_cast<MemoryPhi>(U))
1171 NonOptPhis.insert(PhiUser);
1172
1173 // Replace all our users with our defining access.
1174 What->replaceAllUsesWith(What->getDefiningAccess());
1175
1176 // Let MemorySSA take care of moving it around in the lists.
1177 MSSA->moveTo(What, BB, Where);
1178
1179 // Now reinsert it into the IR and do whatever fixups needed.
1180 if (auto *MD = dyn_cast<MemoryDef>(What))
1181 insertDef(MD, /*RenameUses=*/true);
1182 else
1183 insertUse(cast<MemoryUse>(What), /*RenameUses=*/true);
1184
1185 // Clear dangling pointers. We added all MemoryPhi users, but not all
1186 // of them are removed by fixupDefs().
1187 NonOptPhis.clear();
1188 }
1189
1190 // Move What before Where in the MemorySSA IR.
moveBefore(MemoryUseOrDef * What,MemoryUseOrDef * Where)1191 void MemorySSAUpdater::moveBefore(MemoryUseOrDef *What, MemoryUseOrDef *Where) {
1192 moveTo(What, Where->getBlock(), Where->getIterator());
1193 }
1194
1195 // Move What after Where in the MemorySSA IR.
moveAfter(MemoryUseOrDef * What,MemoryUseOrDef * Where)1196 void MemorySSAUpdater::moveAfter(MemoryUseOrDef *What, MemoryUseOrDef *Where) {
1197 moveTo(What, Where->getBlock(), ++Where->getIterator());
1198 }
1199
moveToPlace(MemoryUseOrDef * What,BasicBlock * BB,MemorySSA::InsertionPlace Where)1200 void MemorySSAUpdater::moveToPlace(MemoryUseOrDef *What, BasicBlock *BB,
1201 MemorySSA::InsertionPlace Where) {
1202 if (Where != MemorySSA::InsertionPlace::BeforeTerminator)
1203 return moveTo(What, BB, Where);
1204
1205 if (auto *Where = MSSA->getMemoryAccess(BB->getTerminator()))
1206 return moveBefore(What, Where);
1207 else
1208 return moveTo(What, BB, MemorySSA::InsertionPlace::End);
1209 }
1210
1211 // All accesses in To used to be in From. Move to end and update access lists.
moveAllAccesses(BasicBlock * From,BasicBlock * To,Instruction * Start)1212 void MemorySSAUpdater::moveAllAccesses(BasicBlock *From, BasicBlock *To,
1213 Instruction *Start) {
1214
1215 MemorySSA::AccessList *Accs = MSSA->getWritableBlockAccesses(From);
1216 if (!Accs)
1217 return;
1218
1219 assert(Start->getParent() == To && "Incorrect Start instruction");
1220 MemoryAccess *FirstInNew = nullptr;
1221 for (Instruction &I : make_range(Start->getIterator(), To->end()))
1222 if ((FirstInNew = MSSA->getMemoryAccess(&I)))
1223 break;
1224 if (FirstInNew) {
1225 auto *MUD = cast<MemoryUseOrDef>(FirstInNew);
1226 do {
1227 auto NextIt = ++MUD->getIterator();
1228 MemoryUseOrDef *NextMUD = (!Accs || NextIt == Accs->end())
1229 ? nullptr
1230 : cast<MemoryUseOrDef>(&*NextIt);
1231 MSSA->moveTo(MUD, To, MemorySSA::End);
1232 // Moving MUD from Accs in the moveTo above, may delete Accs, so we need
1233 // to retrieve it again.
1234 Accs = MSSA->getWritableBlockAccesses(From);
1235 MUD = NextMUD;
1236 } while (MUD);
1237 }
1238
1239 // If all accesses were moved and only a trivial Phi remains, we try to remove
1240 // that Phi. This is needed when From is going to be deleted.
1241 auto *Defs = MSSA->getWritableBlockDefs(From);
1242 if (Defs && !Defs->empty())
1243 if (auto *Phi = dyn_cast<MemoryPhi>(&*Defs->begin()))
1244 tryRemoveTrivialPhi(Phi);
1245 }
1246
moveAllAfterSpliceBlocks(BasicBlock * From,BasicBlock * To,Instruction * Start)1247 void MemorySSAUpdater::moveAllAfterSpliceBlocks(BasicBlock *From,
1248 BasicBlock *To,
1249 Instruction *Start) {
1250 assert(MSSA->getBlockAccesses(To) == nullptr &&
1251 "To block is expected to be free of MemoryAccesses.");
1252 moveAllAccesses(From, To, Start);
1253 for (BasicBlock *Succ : successors(To))
1254 if (MemoryPhi *MPhi = MSSA->getMemoryAccess(Succ))
1255 MPhi->setIncomingBlock(MPhi->getBasicBlockIndex(From), To);
1256 }
1257
moveAllAfterMergeBlocks(BasicBlock * From,BasicBlock * To,Instruction * Start)1258 void MemorySSAUpdater::moveAllAfterMergeBlocks(BasicBlock *From, BasicBlock *To,
1259 Instruction *Start) {
1260 assert(From->getUniquePredecessor() == To &&
1261 "From block is expected to have a single predecessor (To).");
1262 moveAllAccesses(From, To, Start);
1263 for (BasicBlock *Succ : successors(From))
1264 if (MemoryPhi *MPhi = MSSA->getMemoryAccess(Succ))
1265 MPhi->setIncomingBlock(MPhi->getBasicBlockIndex(From), To);
1266 }
1267
wireOldPredecessorsToNewImmediatePredecessor(BasicBlock * Old,BasicBlock * New,ArrayRef<BasicBlock * > Preds,bool IdenticalEdgesWereMerged)1268 void MemorySSAUpdater::wireOldPredecessorsToNewImmediatePredecessor(
1269 BasicBlock *Old, BasicBlock *New, ArrayRef<BasicBlock *> Preds,
1270 bool IdenticalEdgesWereMerged) {
1271 assert(!MSSA->getWritableBlockAccesses(New) &&
1272 "Access list should be null for a new block.");
1273 MemoryPhi *Phi = MSSA->getMemoryAccess(Old);
1274 if (!Phi)
1275 return;
1276 if (Old->hasNPredecessors(1)) {
1277 assert(pred_size(New) == Preds.size() &&
1278 "Should have moved all predecessors.");
1279 MSSA->moveTo(Phi, New, MemorySSA::Beginning);
1280 } else {
1281 assert(!Preds.empty() && "Must be moving at least one predecessor to the "
1282 "new immediate predecessor.");
1283 MemoryPhi *NewPhi = MSSA->createMemoryPhi(New);
1284 SmallPtrSet<BasicBlock *, 16> PredsSet(Preds.begin(), Preds.end());
1285 // Currently only support the case of removing a single incoming edge when
1286 // identical edges were not merged.
1287 if (!IdenticalEdgesWereMerged)
1288 assert(PredsSet.size() == Preds.size() &&
1289 "If identical edges were not merged, we cannot have duplicate "
1290 "blocks in the predecessors");
1291 Phi->unorderedDeleteIncomingIf([&](MemoryAccess *MA, BasicBlock *B) {
1292 if (PredsSet.count(B)) {
1293 NewPhi->addIncoming(MA, B);
1294 if (!IdenticalEdgesWereMerged)
1295 PredsSet.erase(B);
1296 return true;
1297 }
1298 return false;
1299 });
1300 Phi->addIncoming(NewPhi, New);
1301 tryRemoveTrivialPhi(NewPhi);
1302 }
1303 }
1304
removeMemoryAccess(MemoryAccess * MA,bool OptimizePhis)1305 void MemorySSAUpdater::removeMemoryAccess(MemoryAccess *MA, bool OptimizePhis) {
1306 assert(!MSSA->isLiveOnEntryDef(MA) &&
1307 "Trying to remove the live on entry def");
1308 // We can only delete phi nodes if they have no uses, or we can replace all
1309 // uses with a single definition.
1310 MemoryAccess *NewDefTarget = nullptr;
1311 if (MemoryPhi *MP = dyn_cast<MemoryPhi>(MA)) {
1312 // Note that it is sufficient to know that all edges of the phi node have
1313 // the same argument. If they do, by the definition of dominance frontiers
1314 // (which we used to place this phi), that argument must dominate this phi,
1315 // and thus, must dominate the phi's uses, and so we will not hit the assert
1316 // below.
1317 NewDefTarget = onlySingleValue(MP);
1318 assert((NewDefTarget || MP->use_empty()) &&
1319 "We can't delete this memory phi");
1320 } else {
1321 NewDefTarget = cast<MemoryUseOrDef>(MA)->getDefiningAccess();
1322 }
1323
1324 SmallSetVector<MemoryPhi *, 4> PhisToCheck;
1325
1326 // Re-point the uses at our defining access
1327 if (!isa<MemoryUse>(MA) && !MA->use_empty()) {
1328 // Reset optimized on users of this store, and reset the uses.
1329 // A few notes:
1330 // 1. This is a slightly modified version of RAUW to avoid walking the
1331 // uses twice here.
1332 // 2. If we wanted to be complete, we would have to reset the optimized
1333 // flags on users of phi nodes if doing the below makes a phi node have all
1334 // the same arguments. Instead, we prefer users to removeMemoryAccess those
1335 // phi nodes, because doing it here would be N^3.
1336 if (MA->hasValueHandle())
1337 ValueHandleBase::ValueIsRAUWd(MA, NewDefTarget);
1338 // Note: We assume MemorySSA is not used in metadata since it's not really
1339 // part of the IR.
1340
1341 assert(NewDefTarget != MA && "Going into an infinite loop");
1342 while (!MA->use_empty()) {
1343 Use &U = *MA->use_begin();
1344 if (auto *MUD = dyn_cast<MemoryUseOrDef>(U.getUser()))
1345 MUD->resetOptimized();
1346 if (OptimizePhis)
1347 if (MemoryPhi *MP = dyn_cast<MemoryPhi>(U.getUser()))
1348 PhisToCheck.insert(MP);
1349 U.set(NewDefTarget);
1350 }
1351 }
1352
1353 // The call below to erase will destroy MA, so we can't change the order we
1354 // are doing things here
1355 MSSA->removeFromLookups(MA);
1356 MSSA->removeFromLists(MA);
1357
1358 // Optionally optimize Phi uses. This will recursively remove trivial phis.
1359 if (!PhisToCheck.empty()) {
1360 SmallVector<WeakVH, 16> PhisToOptimize{PhisToCheck.begin(),
1361 PhisToCheck.end()};
1362 PhisToCheck.clear();
1363
1364 unsigned PhisSize = PhisToOptimize.size();
1365 while (PhisSize-- > 0)
1366 if (MemoryPhi *MP =
1367 cast_or_null<MemoryPhi>(PhisToOptimize.pop_back_val()))
1368 tryRemoveTrivialPhi(MP);
1369 }
1370 }
1371
removeBlocks(const SmallSetVector<BasicBlock *,8> & DeadBlocks)1372 void MemorySSAUpdater::removeBlocks(
1373 const SmallSetVector<BasicBlock *, 8> &DeadBlocks) {
1374 // First delete all uses of BB in MemoryPhis.
1375 for (BasicBlock *BB : DeadBlocks) {
1376 Instruction *TI = BB->getTerminator();
1377 assert(TI && "Basic block expected to have a terminator instruction");
1378 for (BasicBlock *Succ : successors(TI))
1379 if (!DeadBlocks.count(Succ))
1380 if (MemoryPhi *MP = MSSA->getMemoryAccess(Succ)) {
1381 MP->unorderedDeleteIncomingBlock(BB);
1382 tryRemoveTrivialPhi(MP);
1383 }
1384 // Drop all references of all accesses in BB
1385 if (MemorySSA::AccessList *Acc = MSSA->getWritableBlockAccesses(BB))
1386 for (MemoryAccess &MA : *Acc)
1387 MA.dropAllReferences();
1388 }
1389
1390 // Next, delete all memory accesses in each block
1391 for (BasicBlock *BB : DeadBlocks) {
1392 MemorySSA::AccessList *Acc = MSSA->getWritableBlockAccesses(BB);
1393 if (!Acc)
1394 continue;
1395 for (MemoryAccess &MA : llvm::make_early_inc_range(*Acc)) {
1396 MSSA->removeFromLookups(&MA);
1397 MSSA->removeFromLists(&MA);
1398 }
1399 }
1400 }
1401
tryRemoveTrivialPhis(ArrayRef<WeakVH> UpdatedPHIs)1402 void MemorySSAUpdater::tryRemoveTrivialPhis(ArrayRef<WeakVH> UpdatedPHIs) {
1403 for (auto &VH : UpdatedPHIs)
1404 if (auto *MPhi = cast_or_null<MemoryPhi>(VH))
1405 tryRemoveTrivialPhi(MPhi);
1406 }
1407
changeToUnreachable(const Instruction * I)1408 void MemorySSAUpdater::changeToUnreachable(const Instruction *I) {
1409 const BasicBlock *BB = I->getParent();
1410 // Remove memory accesses in BB for I and all following instructions.
1411 auto BBI = I->getIterator(), BBE = BB->end();
1412 // FIXME: If this becomes too expensive, iterate until the first instruction
1413 // with a memory access, then iterate over MemoryAccesses.
1414 while (BBI != BBE)
1415 removeMemoryAccess(&*(BBI++));
1416 // Update phis in BB's successors to remove BB.
1417 SmallVector<WeakVH, 16> UpdatedPHIs;
1418 for (const BasicBlock *Successor : successors(BB)) {
1419 removeDuplicatePhiEdgesBetween(BB, Successor);
1420 if (MemoryPhi *MPhi = MSSA->getMemoryAccess(Successor)) {
1421 MPhi->unorderedDeleteIncomingBlock(BB);
1422 UpdatedPHIs.push_back(MPhi);
1423 }
1424 }
1425 // Optimize trivial phis.
1426 tryRemoveTrivialPhis(UpdatedPHIs);
1427 }
1428
createMemoryAccessInBB(Instruction * I,MemoryAccess * Definition,const BasicBlock * BB,MemorySSA::InsertionPlace Point)1429 MemoryAccess *MemorySSAUpdater::createMemoryAccessInBB(
1430 Instruction *I, MemoryAccess *Definition, const BasicBlock *BB,
1431 MemorySSA::InsertionPlace Point) {
1432 MemoryUseOrDef *NewAccess = MSSA->createDefinedAccess(I, Definition);
1433 MSSA->insertIntoListsForBlock(NewAccess, BB, Point);
1434 return NewAccess;
1435 }
1436
createMemoryAccessBefore(Instruction * I,MemoryAccess * Definition,MemoryUseOrDef * InsertPt)1437 MemoryUseOrDef *MemorySSAUpdater::createMemoryAccessBefore(
1438 Instruction *I, MemoryAccess *Definition, MemoryUseOrDef *InsertPt) {
1439 assert(I->getParent() == InsertPt->getBlock() &&
1440 "New and old access must be in the same block");
1441 MemoryUseOrDef *NewAccess = MSSA->createDefinedAccess(I, Definition);
1442 MSSA->insertIntoListsBefore(NewAccess, InsertPt->getBlock(),
1443 InsertPt->getIterator());
1444 return NewAccess;
1445 }
1446
createMemoryAccessAfter(Instruction * I,MemoryAccess * Definition,MemoryAccess * InsertPt)1447 MemoryUseOrDef *MemorySSAUpdater::createMemoryAccessAfter(
1448 Instruction *I, MemoryAccess *Definition, MemoryAccess *InsertPt) {
1449 assert(I->getParent() == InsertPt->getBlock() &&
1450 "New and old access must be in the same block");
1451 MemoryUseOrDef *NewAccess = MSSA->createDefinedAccess(I, Definition);
1452 MSSA->insertIntoListsBefore(NewAccess, InsertPt->getBlock(),
1453 ++InsertPt->getIterator());
1454 return NewAccess;
1455 }
1456