1 //===-- Local.cpp - Functions to perform local transformations ------------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This family of functions perform various local transformations to the
11 // program.
12 //
13 //===----------------------------------------------------------------------===//
15 #include "llvm/Transforms/Utils/Local.h"
16 #include "llvm/ADT/DenseMap.h"
17 #include "llvm/ADT/STLExtras.h"
18 #include "llvm/ADT/SmallPtrSet.h"
19 #include "llvm/ADT/Statistic.h"
20 #include "llvm/Analysis/Dominators.h"
21 #include "llvm/Analysis/InstructionSimplify.h"
22 #include "llvm/Analysis/MemoryBuiltins.h"
23 #include "llvm/Analysis/ValueTracking.h"
24 #include "llvm/DIBuilder.h"
25 #include "llvm/DebugInfo.h"
26 #include "llvm/IR/Constants.h"
27 #include "llvm/IR/DataLayout.h"
28 #include "llvm/IR/DerivedTypes.h"
29 #include "llvm/IR/GlobalAlias.h"
30 #include "llvm/IR/GlobalVariable.h"
31 #include "llvm/IR/IRBuilder.h"
32 #include "llvm/IR/Instructions.h"
33 #include "llvm/IR/IntrinsicInst.h"
34 #include "llvm/IR/Intrinsics.h"
35 #include "llvm/IR/MDBuilder.h"
36 #include "llvm/IR/Metadata.h"
37 #include "llvm/IR/Operator.h"
38 #include "llvm/Support/CFG.h"
39 #include "llvm/Support/Debug.h"
40 #include "llvm/Support/GetElementPtrTypeIterator.h"
41 #include "llvm/Support/MathExtras.h"
42 #include "llvm/Support/ValueHandle.h"
43 #include "llvm/Support/raw_ostream.h"
44 using namespace llvm;
46 STATISTIC(NumRemoved, "Number of unreachable basic blocks removed");
48 //===----------------------------------------------------------------------===//
49 // Local constant propagation.
50 //
52 /// ConstantFoldTerminator - If a terminator instruction is predicated on a
53 /// constant value, convert it into an unconditional branch to the constant
54 /// destination. This is a nontrivial operation because the successors of this
55 /// basic block must have their PHI nodes updated.
56 /// Also calls RecursivelyDeleteTriviallyDeadInstructions() on any branch/switch
57 /// conditions and indirectbr addresses this might make dead if
58 /// DeleteDeadConditions is true.
59 bool llvm::ConstantFoldTerminator(BasicBlock *BB, bool DeleteDeadConditions,
60 const TargetLibraryInfo *TLI) {
61 TerminatorInst *T = BB->getTerminator();
62 IRBuilder<> Builder(T);
64 // Branch - See if we are conditional jumping on constant
65 if (BranchInst *BI = dyn_cast<BranchInst>(T)) {
66 if (BI->isUnconditional()) return false; // Can't optimize uncond branch
67 BasicBlock *Dest1 = BI->getSuccessor(0);
68 BasicBlock *Dest2 = BI->getSuccessor(1);
70 if (ConstantInt *Cond = dyn_cast<ConstantInt>(BI->getCondition())) {
71 // Are we branching on constant?
72 // YES. Change to unconditional branch...
73 BasicBlock *Destination = Cond->getZExtValue() ? Dest1 : Dest2;
74 BasicBlock *OldDest = Cond->getZExtValue() ? Dest2 : Dest1;
76 //cerr << "Function: " << T->getParent()->getParent()
77 // << "\nRemoving branch from " << T->getParent()
78 // << "\n\nTo: " << OldDest << endl;
80 // Let the basic block know that we are letting go of it. Based on this,
81 // it will adjust it's PHI nodes.
82 OldDest->removePredecessor(BB);
84 // Replace the conditional branch with an unconditional one.
85 Builder.CreateBr(Destination);
86 BI->eraseFromParent();
87 return true;
88 }
90 if (Dest2 == Dest1) { // Conditional branch to same location?
91 // This branch matches something like this:
92 // br bool %cond, label %Dest, label %Dest
93 // and changes it into: br label %Dest
95 // Let the basic block know that we are letting go of one copy of it.
96 assert(BI->getParent() && "Terminator not inserted in block!");
97 Dest1->removePredecessor(BI->getParent());
99 // Replace the conditional branch with an unconditional one.
100 Builder.CreateBr(Dest1);
101 Value *Cond = BI->getCondition();
102 BI->eraseFromParent();
103 if (DeleteDeadConditions)
104 RecursivelyDeleteTriviallyDeadInstructions(Cond, TLI);
105 return true;
106 }
107 return false;
108 }
110 if (SwitchInst *SI = dyn_cast<SwitchInst>(T)) {
111 // If we are switching on a constant, we can convert the switch into a
112 // single branch instruction!
113 ConstantInt *CI = dyn_cast<ConstantInt>(SI->getCondition());
114 BasicBlock *TheOnlyDest = SI->getDefaultDest();
115 BasicBlock *DefaultDest = TheOnlyDest;
117 // Figure out which case it goes to.
118 for (SwitchInst::CaseIt i = SI->case_begin(), e = SI->case_end();
119 i != e; ++i) {
120 // Found case matching a constant operand?
121 if (i.getCaseValue() == CI) {
122 TheOnlyDest = i.getCaseSuccessor();
123 break;
124 }
126 // Check to see if this branch is going to the same place as the default
127 // dest. If so, eliminate it as an explicit compare.
128 if (i.getCaseSuccessor() == DefaultDest) {
129 MDNode* MD = SI->getMetadata(LLVMContext::MD_prof);
130 // MD should have 2 + NumCases operands.
131 if (MD && MD->getNumOperands() == 2 + SI->getNumCases()) {
132 // Collect branch weights into a vector.
133 SmallVector<uint32_t, 8> Weights;
134 for (unsigned MD_i = 1, MD_e = MD->getNumOperands(); MD_i < MD_e;
135 ++MD_i) {
136 ConstantInt* CI = dyn_cast<ConstantInt>(MD->getOperand(MD_i));
137 assert(CI);
138 Weights.push_back(CI->getValue().getZExtValue());
139 }
140 // Merge weight of this case to the default weight.
141 unsigned idx = i.getCaseIndex();
142 Weights[0] += Weights[idx+1];
143 // Remove weight for this case.
144 std::swap(Weights[idx+1], Weights.back());
145 Weights.pop_back();
146 SI->setMetadata(LLVMContext::MD_prof,
147 MDBuilder(BB->getContext()).
148 createBranchWeights(Weights));
149 }
150 // Remove this entry.
151 DefaultDest->removePredecessor(SI->getParent());
152 SI->removeCase(i);
153 --i; --e;
154 continue;
155 }
157 // Otherwise, check to see if the switch only branches to one destination.
158 // We do this by reseting "TheOnlyDest" to null when we find two non-equal
159 // destinations.
160 if (i.getCaseSuccessor() != TheOnlyDest) TheOnlyDest = 0;
161 }
163 if (CI && !TheOnlyDest) {
164 // Branching on a constant, but not any of the cases, go to the default
165 // successor.
166 TheOnlyDest = SI->getDefaultDest();
167 }
169 // If we found a single destination that we can fold the switch into, do so
170 // now.
171 if (TheOnlyDest) {
172 // Insert the new branch.
173 Builder.CreateBr(TheOnlyDest);
174 BasicBlock *BB = SI->getParent();
176 // Remove entries from PHI nodes which we no longer branch to...
177 for (unsigned i = 0, e = SI->getNumSuccessors(); i != e; ++i) {
178 // Found case matching a constant operand?
179 BasicBlock *Succ = SI->getSuccessor(i);
180 if (Succ == TheOnlyDest)
181 TheOnlyDest = 0; // Don't modify the first branch to TheOnlyDest
182 else
183 Succ->removePredecessor(BB);
184 }
186 // Delete the old switch.
187 Value *Cond = SI->getCondition();
188 SI->eraseFromParent();
189 if (DeleteDeadConditions)
190 RecursivelyDeleteTriviallyDeadInstructions(Cond, TLI);
191 return true;
192 }
194 if (SI->getNumCases() == 1) {
195 // Otherwise, we can fold this switch into a conditional branch
196 // instruction if it has only one non-default destination.
197 SwitchInst::CaseIt FirstCase = SI->case_begin();
198 Value *Cond = Builder.CreateICmpEQ(SI->getCondition(),
199 FirstCase.getCaseValue(), "cond");
201 // Insert the new branch.
202 BranchInst *NewBr = Builder.CreateCondBr(Cond,
203 FirstCase.getCaseSuccessor(),
204 SI->getDefaultDest());
205 MDNode* MD = SI->getMetadata(LLVMContext::MD_prof);
206 if (MD && MD->getNumOperands() == 3) {
207 ConstantInt *SICase = dyn_cast<ConstantInt>(MD->getOperand(2));
208 ConstantInt *SIDef = dyn_cast<ConstantInt>(MD->getOperand(1));
209 assert(SICase && SIDef);
210 // The TrueWeight should be the weight for the single case of SI.
211 NewBr->setMetadata(LLVMContext::MD_prof,
212 MDBuilder(BB->getContext()).
213 createBranchWeights(SICase->getValue().getZExtValue(),
214 SIDef->getValue().getZExtValue()));
215 }
217 // Delete the old switch.
218 SI->eraseFromParent();
219 return true;
220 }
221 return false;
222 }
224 if (IndirectBrInst *IBI = dyn_cast<IndirectBrInst>(T)) {
225 // indirectbr blockaddress(@F, @BB) -> br label @BB
226 if (BlockAddress *BA =
227 dyn_cast<BlockAddress>(IBI->getAddress()->stripPointerCasts())) {
228 BasicBlock *TheOnlyDest = BA->getBasicBlock();
229 // Insert the new branch.
230 Builder.CreateBr(TheOnlyDest);
232 for (unsigned i = 0, e = IBI->getNumDestinations(); i != e; ++i) {
233 if (IBI->getDestination(i) == TheOnlyDest)
234 TheOnlyDest = 0;
235 else
236 IBI->getDestination(i)->removePredecessor(IBI->getParent());
237 }
238 Value *Address = IBI->getAddress();
239 IBI->eraseFromParent();
240 if (DeleteDeadConditions)
241 RecursivelyDeleteTriviallyDeadInstructions(Address, TLI);
243 // If we didn't find our destination in the IBI successor list, then we
244 // have undefined behavior. Replace the unconditional branch with an
245 // 'unreachable' instruction.
246 if (TheOnlyDest) {
247 BB->getTerminator()->eraseFromParent();
248 new UnreachableInst(BB->getContext(), BB);
249 }
251 return true;
252 }
253 }
255 return false;
256 }
259 //===----------------------------------------------------------------------===//
260 // Local dead code elimination.
261 //
263 /// isInstructionTriviallyDead - Return true if the result produced by the
264 /// instruction is not used, and the instruction has no side effects.
265 ///
266 bool llvm::isInstructionTriviallyDead(Instruction *I,
267 const TargetLibraryInfo *TLI) {
268 if (!I->use_empty() || isa<TerminatorInst>(I)) return false;
270 // We don't want the landingpad instruction removed by anything this general.
271 if (isa<LandingPadInst>(I))
272 return false;
274 // We don't want debug info removed by anything this general, unless
275 // debug info is empty.
276 if (DbgDeclareInst *DDI = dyn_cast<DbgDeclareInst>(I)) {
277 if (DDI->getAddress())
278 return false;
279 return true;
280 }
281 if (DbgValueInst *DVI = dyn_cast<DbgValueInst>(I)) {
282 if (DVI->getValue())
283 return false;
284 return true;
285 }
287 if (!I->mayHaveSideEffects()) return true;
289 // Special case intrinsics that "may have side effects" but can be deleted
290 // when dead.
291 if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(I)) {
292 // Safe to delete llvm.stacksave if dead.
293 if (II->getIntrinsicID() == Intrinsic::stacksave)
294 return true;
296 // Lifetime intrinsics are dead when their right-hand is undef.
297 if (II->getIntrinsicID() == Intrinsic::lifetime_start ||
298 II->getIntrinsicID() == Intrinsic::lifetime_end)
299 return isa<UndefValue>(II->getArgOperand(1));
300 }
302 if (isAllocLikeFn(I, TLI)) return true;
304 if (CallInst *CI = isFreeCall(I, TLI))
305 if (Constant *C = dyn_cast<Constant>(CI->getArgOperand(0)))
306 return C->isNullValue() || isa<UndefValue>(C);
308 return false;
309 }
311 /// RecursivelyDeleteTriviallyDeadInstructions - If the specified value is a
312 /// trivially dead instruction, delete it. If that makes any of its operands
313 /// trivially dead, delete them too, recursively. Return true if any
314 /// instructions were deleted.
315 bool
316 llvm::RecursivelyDeleteTriviallyDeadInstructions(Value *V,
317 const TargetLibraryInfo *TLI) {
318 Instruction *I = dyn_cast<Instruction>(V);
319 if (!I || !I->use_empty() || !isInstructionTriviallyDead(I, TLI))
320 return false;
322 SmallVector<Instruction*, 16> DeadInsts;
323 DeadInsts.push_back(I);
325 do {
326 I = DeadInsts.pop_back_val();
328 // Null out all of the instruction's operands to see if any operand becomes
329 // dead as we go.
330 for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i) {
331 Value *OpV = I->getOperand(i);
332 I->setOperand(i, 0);
334 if (!OpV->use_empty()) continue;
336 // If the operand is an instruction that became dead as we nulled out the
337 // operand, and if it is 'trivially' dead, delete it in a future loop
338 // iteration.
339 if (Instruction *OpI = dyn_cast<Instruction>(OpV))
340 if (isInstructionTriviallyDead(OpI, TLI))
341 DeadInsts.push_back(OpI);
342 }
344 I->eraseFromParent();
345 } while (!DeadInsts.empty());
347 return true;
348 }
350 /// areAllUsesEqual - Check whether the uses of a value are all the same.
351 /// This is similar to Instruction::hasOneUse() except this will also return
352 /// true when there are no uses or multiple uses that all refer to the same
353 /// value.
354 static bool areAllUsesEqual(Instruction *I) {
355 Value::use_iterator UI = I->use_begin();
356 Value::use_iterator UE = I->use_end();
357 if (UI == UE)
358 return true;
360 User *TheUse = *UI;
361 for (++UI; UI != UE; ++UI) {
362 if (*UI != TheUse)
363 return false;
364 }
365 return true;
366 }
368 /// RecursivelyDeleteDeadPHINode - If the specified value is an effectively
369 /// dead PHI node, due to being a def-use chain of single-use nodes that
370 /// either forms a cycle or is terminated by a trivially dead instruction,
371 /// delete it. If that makes any of its operands trivially dead, delete them
372 /// too, recursively. Return true if a change was made.
373 bool llvm::RecursivelyDeleteDeadPHINode(PHINode *PN,
374 const TargetLibraryInfo *TLI) {
375 SmallPtrSet<Instruction*, 4> Visited;
376 for (Instruction *I = PN; areAllUsesEqual(I) && !I->mayHaveSideEffects();
377 I = cast<Instruction>(*I->use_begin())) {
378 if (I->use_empty())
379 return RecursivelyDeleteTriviallyDeadInstructions(I, TLI);
381 // If we find an instruction more than once, we're on a cycle that
382 // won't prove fruitful.
383 if (!Visited.insert(I)) {
384 // Break the cycle and delete the instruction and its operands.
385 I->replaceAllUsesWith(UndefValue::get(I->getType()));
386 (void)RecursivelyDeleteTriviallyDeadInstructions(I, TLI);
387 return true;
388 }
389 }
390 return false;
391 }
393 /// SimplifyInstructionsInBlock - Scan the specified basic block and try to
394 /// simplify any instructions in it and recursively delete dead instructions.
395 ///
396 /// This returns true if it changed the code, note that it can delete
397 /// instructions in other blocks as well in this block.
398 bool llvm::SimplifyInstructionsInBlock(BasicBlock *BB, const DataLayout *TD,
399 const TargetLibraryInfo *TLI) {
400 bool MadeChange = false;
402 #ifndef NDEBUG
403 // In debug builds, ensure that the terminator of the block is never replaced
404 // or deleted by these simplifications. The idea of simplification is that it
405 // cannot introduce new instructions, and there is no way to replace the
406 // terminator of a block without introducing a new instruction.
407 AssertingVH<Instruction> TerminatorVH(--BB->end());
408 #endif
410 for (BasicBlock::iterator BI = BB->begin(), E = --BB->end(); BI != E; ) {
411 assert(!BI->isTerminator());
412 Instruction *Inst = BI++;
414 WeakVH BIHandle(BI);
415 if (recursivelySimplifyInstruction(Inst, TD, TLI)) {
416 MadeChange = true;
417 if (BIHandle != BI)
418 BI = BB->begin();
419 continue;
420 }
422 MadeChange |= RecursivelyDeleteTriviallyDeadInstructions(Inst, TLI);
423 if (BIHandle != BI)
424 BI = BB->begin();
425 }
426 return MadeChange;
427 }
429 //===----------------------------------------------------------------------===//
430 // Control Flow Graph Restructuring.
431 //
434 /// RemovePredecessorAndSimplify - Like BasicBlock::removePredecessor, this
435 /// method is called when we're about to delete Pred as a predecessor of BB. If
436 /// BB contains any PHI nodes, this drops the entries in the PHI nodes for Pred.
437 ///
438 /// Unlike the removePredecessor method, this attempts to simplify uses of PHI
439 /// nodes that collapse into identity values. For example, if we have:
440 /// x = phi(1, 0, 0, 0)
441 /// y = and x, z
442 ///
443 /// .. and delete the predecessor corresponding to the '1', this will attempt to
444 /// recursively fold the and to 0.
445 void llvm::RemovePredecessorAndSimplify(BasicBlock *BB, BasicBlock *Pred,
446 DataLayout *TD) {
447 // This only adjusts blocks with PHI nodes.
448 if (!isa<PHINode>(BB->begin()))
449 return;
451 // Remove the entries for Pred from the PHI nodes in BB, but do not simplify
452 // them down. This will leave us with single entry phi nodes and other phis
453 // that can be removed.
454 BB->removePredecessor(Pred, true);
456 WeakVH PhiIt = &BB->front();
457 while (PHINode *PN = dyn_cast<PHINode>(PhiIt)) {
458 PhiIt = &*++BasicBlock::iterator(cast<Instruction>(PhiIt));
459 Value *OldPhiIt = PhiIt;
461 if (!recursivelySimplifyInstruction(PN, TD))
462 continue;
464 // If recursive simplification ended up deleting the next PHI node we would
465 // iterate to, then our iterator is invalid, restart scanning from the top
466 // of the block.
467 if (PhiIt != OldPhiIt) PhiIt = &BB->front();
468 }
469 }
472 /// MergeBasicBlockIntoOnlyPred - DestBB is a block with one predecessor and its
473 /// predecessor is known to have one successor (DestBB!). Eliminate the edge
474 /// between them, moving the instructions in the predecessor into DestBB and
475 /// deleting the predecessor block.
476 ///
477 void llvm::MergeBasicBlockIntoOnlyPred(BasicBlock *DestBB, Pass *P) {
478 // If BB has single-entry PHI nodes, fold them.
479 while (PHINode *PN = dyn_cast<PHINode>(DestBB->begin())) {
480 Value *NewVal = PN->getIncomingValue(0);
481 // Replace self referencing PHI with undef, it must be dead.
482 if (NewVal == PN) NewVal = UndefValue::get(PN->getType());
483 PN->replaceAllUsesWith(NewVal);
484 PN->eraseFromParent();
485 }
487 BasicBlock *PredBB = DestBB->getSinglePredecessor();
488 assert(PredBB && "Block doesn't have a single predecessor!");
490 // Zap anything that took the address of DestBB. Not doing this will give the
491 // address an invalid value.
492 if (DestBB->hasAddressTaken()) {
493 BlockAddress *BA = BlockAddress::get(DestBB);
494 Constant *Replacement =
495 ConstantInt::get(llvm::Type::getInt32Ty(BA->getContext()), 1);
496 BA->replaceAllUsesWith(ConstantExpr::getIntToPtr(Replacement,
497 BA->getType()));
498 BA->destroyConstant();
499 }
501 // Anything that branched to PredBB now branches to DestBB.
502 PredBB->replaceAllUsesWith(DestBB);
504 // Splice all the instructions from PredBB to DestBB.
505 PredBB->getTerminator()->eraseFromParent();
506 DestBB->getInstList().splice(DestBB->begin(), PredBB->getInstList());
508 if (P) {
509 DominatorTree *DT = P->getAnalysisIfAvailable<DominatorTree>();
510 if (DT) {
511 BasicBlock *PredBBIDom = DT->getNode(PredBB)->getIDom()->getBlock();
512 DT->changeImmediateDominator(DestBB, PredBBIDom);
513 DT->eraseNode(PredBB);
514 }
515 }
516 // Nuke BB.
517 PredBB->eraseFromParent();
518 }
520 /// CanMergeValues - Return true if we can choose one of these values to use
521 /// in place of the other. Note that we will always choose the non-undef
522 /// value to keep.
523 static bool CanMergeValues(Value *First, Value *Second) {
524 return First == Second || isa<UndefValue>(First) || isa<UndefValue>(Second);
525 }
527 /// CanPropagatePredecessorsForPHIs - Return true if we can fold BB, an
528 /// almost-empty BB ending in an unconditional branch to Succ, into Succ.
529 ///
530 /// Assumption: Succ is the single successor for BB.
531 ///
532 static bool CanPropagatePredecessorsForPHIs(BasicBlock *BB, BasicBlock *Succ) {
533 assert(*succ_begin(BB) == Succ && "Succ is not successor of BB!");
535 DEBUG(dbgs() << "Looking to fold " << BB->getName() << " into "
536 << Succ->getName() << "\n");
537 // Shortcut, if there is only a single predecessor it must be BB and merging
538 // is always safe
539 if (Succ->getSinglePredecessor()) return true;
541 // Make a list of the predecessors of BB
542 SmallPtrSet<BasicBlock*, 16> BBPreds(pred_begin(BB), pred_end(BB));
544 // Look at all the phi nodes in Succ, to see if they present a conflict when
545 // merging these blocks
546 for (BasicBlock::iterator I = Succ->begin(); isa<PHINode>(I); ++I) {
547 PHINode *PN = cast<PHINode>(I);
549 // If the incoming value from BB is again a PHINode in
550 // BB which has the same incoming value for *PI as PN does, we can
551 // merge the phi nodes and then the blocks can still be merged
552 PHINode *BBPN = dyn_cast<PHINode>(PN->getIncomingValueForBlock(BB));
553 if (BBPN && BBPN->getParent() == BB) {
554 for (unsigned PI = 0, PE = PN->getNumIncomingValues(); PI != PE; ++PI) {
555 BasicBlock *IBB = PN->getIncomingBlock(PI);
556 if (BBPreds.count(IBB) &&
557 !CanMergeValues(BBPN->getIncomingValueForBlock(IBB),
558 PN->getIncomingValue(PI))) {
559 DEBUG(dbgs() << "Can't fold, phi node " << PN->getName() << " in "
560 << Succ->getName() << " is conflicting with "
561 << BBPN->getName() << " with regard to common predecessor "
562 << IBB->getName() << "\n");
563 return false;
564 }
565 }
566 } else {
567 Value* Val = PN->getIncomingValueForBlock(BB);
568 for (unsigned PI = 0, PE = PN->getNumIncomingValues(); PI != PE; ++PI) {
569 // See if the incoming value for the common predecessor is equal to the
570 // one for BB, in which case this phi node will not prevent the merging
571 // of the block.
572 BasicBlock *IBB = PN->getIncomingBlock(PI);
573 if (BBPreds.count(IBB) &&
574 !CanMergeValues(Val, PN->getIncomingValue(PI))) {
575 DEBUG(dbgs() << "Can't fold, phi node " << PN->getName() << " in "
576 << Succ->getName() << " is conflicting with regard to common "
577 << "predecessor " << IBB->getName() << "\n");
578 return false;
579 }
580 }
581 }
582 }
584 return true;
585 }
587 typedef SmallVector<BasicBlock *, 16> PredBlockVector;
588 typedef DenseMap<BasicBlock *, Value *> IncomingValueMap;
590 /// \brief Determines the value to use as the phi node input for a block.
591 ///
592 /// Select between \p OldVal any value that we know flows from \p BB
593 /// to a particular phi on the basis of which one (if either) is not
594 /// undef. Update IncomingValues based on the selected value.
595 ///
596 /// \param OldVal The value we are considering selecting.
597 /// \param BB The block that the value flows in from.
598 /// \param IncomingValues A map from block-to-value for other phi inputs
599 /// that we have examined.
600 ///
601 /// \returns the selected value.
602 static Value *selectIncomingValueForBlock(Value *OldVal, BasicBlock *BB,
603 IncomingValueMap &IncomingValues) {
604 if (!isa<UndefValue>(OldVal)) {
605 assert((!IncomingValues.count(BB) ||
606 IncomingValues.find(BB)->second == OldVal) &&
607 "Expected OldVal to match incoming value from BB!");
609 IncomingValues.insert(std::make_pair(BB, OldVal));
610 return OldVal;
611 }
613 IncomingValueMap::const_iterator It = IncomingValues.find(BB);
614 if (It != IncomingValues.end()) return It->second;
616 return OldVal;
617 }
619 /// \brief Create a map from block to value for the operands of a
620 /// given phi.
621 ///
622 /// Create a map from block to value for each non-undef value flowing
623 /// into \p PN.
624 ///
625 /// \param PN The phi we are collecting the map for.
626 /// \param IncomingValues [out] The map from block to value for this phi.
627 static void gatherIncomingValuesToPhi(PHINode *PN,
628 IncomingValueMap &IncomingValues) {
629 for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
630 BasicBlock *BB = PN->getIncomingBlock(i);
631 Value *V = PN->getIncomingValue(i);
633 if (!isa<UndefValue>(V))
634 IncomingValues.insert(std::make_pair(BB, V));
635 }
636 }
638 /// \brief Replace the incoming undef values to a phi with the values
639 /// from a block-to-value map.
640 ///
641 /// \param PN The phi we are replacing the undefs in.
642 /// \param IncomingValues A map from block to value.
643 static void replaceUndefValuesInPhi(PHINode *PN,
644 const IncomingValueMap &IncomingValues) {
645 for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
646 Value *V = PN->getIncomingValue(i);
648 if (!isa<UndefValue>(V)) continue;
650 BasicBlock *BB = PN->getIncomingBlock(i);
651 IncomingValueMap::const_iterator It = IncomingValues.find(BB);
652 if (It == IncomingValues.end()) continue;
654 PN->setIncomingValue(i, It->second);
655 }
656 }
658 /// \brief Replace a value flowing from a block to a phi with
659 /// potentially multiple instances of that value flowing from the
660 /// block's predecessors to the phi.
661 ///
662 /// \param BB The block with the value flowing into the phi.
663 /// \param BBPreds The predecessors of BB.
664 /// \param PN The phi that we are updating.
665 static void redirectValuesFromPredecessorsToPhi(BasicBlock *BB,
666 const PredBlockVector &BBPreds,
667 PHINode *PN) {
668 Value *OldVal = PN->removeIncomingValue(BB, false);
669 assert(OldVal && "No entry in PHI for Pred BB!");
671 IncomingValueMap IncomingValues;
673 // We are merging two blocks - BB, and the block containing PN - and
674 // as a result we need to redirect edges from the predecessors of BB
675 // to go to the block containing PN, and update PN
676 // accordingly. Since we allow merging blocks in the case where the
677 // predecessor and successor blocks both share some predecessors,
678 // and where some of those common predecessors might have undef
679 // values flowing into PN, we want to rewrite those values to be
680 // consistent with the non-undef values.
682 gatherIncomingValuesToPhi(PN, IncomingValues);
684 // If this incoming value is one of the PHI nodes in BB, the new entries
685 // in the PHI node are the entries from the old PHI.
686 if (isa<PHINode>(OldVal) && cast<PHINode>(OldVal)->getParent() == BB) {
687 PHINode *OldValPN = cast<PHINode>(OldVal);
688 for (unsigned i = 0, e = OldValPN->getNumIncomingValues(); i != e; ++i) {
689 // Note that, since we are merging phi nodes and BB and Succ might
690 // have common predecessors, we could end up with a phi node with
691 // identical incoming branches. This will be cleaned up later (and
692 // will trigger asserts if we try to clean it up now, without also
693 // simplifying the corresponding conditional branch).
694 BasicBlock *PredBB = OldValPN->getIncomingBlock(i);
695 Value *PredVal = OldValPN->getIncomingValue(i);
696 Value *Selected = selectIncomingValueForBlock(PredVal, PredBB,
697 IncomingValues);
699 // And add a new incoming value for this predecessor for the
700 // newly retargeted branch.
701 PN->addIncoming(Selected, PredBB);
702 }
703 } else {
704 for (unsigned i = 0, e = BBPreds.size(); i != e; ++i) {
705 // Update existing incoming values in PN for this
706 // predecessor of BB.
707 BasicBlock *PredBB = BBPreds[i];
708 Value *Selected = selectIncomingValueForBlock(OldVal, PredBB,
709 IncomingValues);
711 // And add a new incoming value for this predecessor for the
712 // newly retargeted branch.
713 PN->addIncoming(Selected, PredBB);
714 }
715 }
717 replaceUndefValuesInPhi(PN, IncomingValues);
718 }
720 /// TryToSimplifyUncondBranchFromEmptyBlock - BB is known to contain an
721 /// unconditional branch, and contains no instructions other than PHI nodes,
722 /// potential side-effect free intrinsics and the branch. If possible,
723 /// eliminate BB by rewriting all the predecessors to branch to the successor
724 /// block and return true. If we can't transform, return false.
725 bool llvm::TryToSimplifyUncondBranchFromEmptyBlock(BasicBlock *BB) {
726 assert(BB != &BB->getParent()->getEntryBlock() &&
727 "TryToSimplifyUncondBranchFromEmptyBlock called on entry block!");
729 // We can't eliminate infinite loops.
730 BasicBlock *Succ = cast<BranchInst>(BB->getTerminator())->getSuccessor(0);
731 if (BB == Succ) return false;
733 // Check to see if merging these blocks would cause conflicts for any of the
734 // phi nodes in BB or Succ. If not, we can safely merge.
735 if (!CanPropagatePredecessorsForPHIs(BB, Succ)) return false;
737 // Check for cases where Succ has multiple predecessors and a PHI node in BB
738 // has uses which will not disappear when the PHI nodes are merged. It is
739 // possible to handle such cases, but difficult: it requires checking whether
740 // BB dominates Succ, which is non-trivial to calculate in the case where
741 // Succ has multiple predecessors. Also, it requires checking whether
742 // constructing the necessary self-referential PHI node doesn't introduce any
743 // conflicts; this isn't too difficult, but the previous code for doing this
744 // was incorrect.
745 //
746 // Note that if this check finds a live use, BB dominates Succ, so BB is
747 // something like a loop pre-header (or rarely, a part of an irreducible CFG);
748 // folding the branch isn't profitable in that case anyway.
749 if (!Succ->getSinglePredecessor()) {
750 BasicBlock::iterator BBI = BB->begin();
751 while (isa<PHINode>(*BBI)) {
752 for (Value::use_iterator UI = BBI->use_begin(), E = BBI->use_end();
753 UI != E; ++UI) {
754 if (PHINode* PN = dyn_cast<PHINode>(*UI)) {
755 if (PN->getIncomingBlock(UI) != BB)
756 return false;
757 } else {
758 return false;
759 }
760 }
761 ++BBI;
762 }
763 }
765 DEBUG(dbgs() << "Killing Trivial BB: \n" << *BB);
767 if (isa<PHINode>(Succ->begin())) {
768 // If there is more than one pred of succ, and there are PHI nodes in
769 // the successor, then we need to add incoming edges for the PHI nodes
770 //
771 const PredBlockVector BBPreds(pred_begin(BB), pred_end(BB));
773 // Loop over all of the PHI nodes in the successor of BB.
774 for (BasicBlock::iterator I = Succ->begin(); isa<PHINode>(I); ++I) {
775 PHINode *PN = cast<PHINode>(I);
777 redirectValuesFromPredecessorsToPhi(BB, BBPreds, PN);
778 }
779 }
781 if (Succ->getSinglePredecessor()) {
782 // BB is the only predecessor of Succ, so Succ will end up with exactly
783 // the same predecessors BB had.
785 // Copy over any phi, debug or lifetime instruction.
786 BB->getTerminator()->eraseFromParent();
787 Succ->getInstList().splice(Succ->getFirstNonPHI(), BB->getInstList());
788 } else {
789 while (PHINode *PN = dyn_cast<PHINode>(&BB->front())) {
790 // We explicitly check for such uses in CanPropagatePredecessorsForPHIs.
791 assert(PN->use_empty() && "There shouldn't be any uses here!");
792 PN->eraseFromParent();
793 }
794 }
796 // Everything that jumped to BB now goes to Succ.
797 BB->replaceAllUsesWith(Succ);
798 if (!Succ->hasName()) Succ->takeName(BB);
799 BB->eraseFromParent(); // Delete the old basic block.
800 return true;
801 }
803 /// EliminateDuplicatePHINodes - Check for and eliminate duplicate PHI
804 /// nodes in this block. This doesn't try to be clever about PHI nodes
805 /// which differ only in the order of the incoming values, but instcombine
806 /// orders them so it usually won't matter.
807 ///
808 bool llvm::EliminateDuplicatePHINodes(BasicBlock *BB) {
809 bool Changed = false;
811 // This implementation doesn't currently consider undef operands
812 // specially. Theoretically, two phis which are identical except for
813 // one having an undef where the other doesn't could be collapsed.
815 // Map from PHI hash values to PHI nodes. If multiple PHIs have
816 // the same hash value, the element is the first PHI in the
817 // linked list in CollisionMap.
818 DenseMap<uintptr_t, PHINode *> HashMap;
820 // Maintain linked lists of PHI nodes with common hash values.
821 DenseMap<PHINode *, PHINode *> CollisionMap;
823 // Examine each PHI.
824 for (BasicBlock::iterator I = BB->begin();
825 PHINode *PN = dyn_cast<PHINode>(I++); ) {
826 // Compute a hash value on the operands. Instcombine will likely have sorted
827 // them, which helps expose duplicates, but we have to check all the
828 // operands to be safe in case instcombine hasn't run.
829 uintptr_t Hash = 0;
830 // This hash algorithm is quite weak as hash functions go, but it seems
831 // to do a good enough job for this particular purpose, and is very quick.
832 for (User::op_iterator I = PN->op_begin(), E = PN->op_end(); I != E; ++I) {
833 Hash ^= reinterpret_cast<uintptr_t>(static_cast<Value *>(*I));
834 Hash = (Hash << 7) | (Hash >> (sizeof(uintptr_t) * CHAR_BIT - 7));
835 }
836 for (PHINode::block_iterator I = PN->block_begin(), E = PN->block_end();
837 I != E; ++I) {
838 Hash ^= reinterpret_cast<uintptr_t>(static_cast<BasicBlock *>(*I));
839 Hash = (Hash << 7) | (Hash >> (sizeof(uintptr_t) * CHAR_BIT - 7));
840 }
841 // Avoid colliding with the DenseMap sentinels ~0 and ~0-1.
842 Hash >>= 1;
843 // If we've never seen this hash value before, it's a unique PHI.
844 std::pair<DenseMap<uintptr_t, PHINode *>::iterator, bool> Pair =
845 HashMap.insert(std::make_pair(Hash, PN));
846 if (Pair.second) continue;
847 // Otherwise it's either a duplicate or a hash collision.
848 for (PHINode *OtherPN = Pair.first->second; ; ) {
849 if (OtherPN->isIdenticalTo(PN)) {
850 // A duplicate. Replace this PHI with its duplicate.
851 PN->replaceAllUsesWith(OtherPN);
852 PN->eraseFromParent();
853 Changed = true;
854 break;
855 }
856 // A non-duplicate hash collision.
857 DenseMap<PHINode *, PHINode *>::iterator I = CollisionMap.find(OtherPN);
858 if (I == CollisionMap.end()) {
859 // Set this PHI to be the head of the linked list of colliding PHIs.
860 PHINode *Old = Pair.first->second;
861 Pair.first->second = PN;
862 CollisionMap[PN] = Old;
863 break;
864 }
865 // Proceed to the next PHI in the list.
866 OtherPN = I->second;
867 }
868 }
870 return Changed;
871 }
873 /// enforceKnownAlignment - If the specified pointer points to an object that
874 /// we control, modify the object's alignment to PrefAlign. This isn't
875 /// often possible though. If alignment is important, a more reliable approach
876 /// is to simply align all global variables and allocation instructions to
877 /// their preferred alignment from the beginning.
878 ///
879 static unsigned enforceKnownAlignment(Value *V, unsigned Align,
880 unsigned PrefAlign, const DataLayout *TD) {
881 V = V->stripPointerCasts();
883 if (AllocaInst *AI = dyn_cast<AllocaInst>(V)) {
884 // If the preferred alignment is greater than the natural stack alignment
885 // then don't round up. This avoids dynamic stack realignment.
886 if (TD && TD->exceedsNaturalStackAlignment(PrefAlign))
887 return Align;
888 // If there is a requested alignment and if this is an alloca, round up.
889 if (AI->getAlignment() >= PrefAlign)
890 return AI->getAlignment();
891 AI->setAlignment(PrefAlign);
892 return PrefAlign;
893 }
895 if (GlobalValue *GV = dyn_cast<GlobalValue>(V)) {
896 // If there is a large requested alignment and we can, bump up the alignment
897 // of the global.
898 if (GV->isDeclaration()) return Align;
899 // If the memory we set aside for the global may not be the memory used by
900 // the final program then it is impossible for us to reliably enforce the
901 // preferred alignment.
902 if (GV->isWeakForLinker()) return Align;
904 if (GV->getAlignment() >= PrefAlign)
905 return GV->getAlignment();
906 // We can only increase the alignment of the global if it has no alignment
907 // specified or if it is not assigned a section. If it is assigned a
908 // section, the global could be densely packed with other objects in the
909 // section, increasing the alignment could cause padding issues.
910 if (!GV->hasSection() || GV->getAlignment() == 0)
911 GV->setAlignment(PrefAlign);
912 return GV->getAlignment();
913 }
915 return Align;
916 }
918 /// getOrEnforceKnownAlignment - If the specified pointer has an alignment that
919 /// we can determine, return it, otherwise return 0. If PrefAlign is specified,
920 /// and it is more than the alignment of the ultimate object, see if we can
921 /// increase the alignment of the ultimate object, making this check succeed.
922 unsigned llvm::getOrEnforceKnownAlignment(Value *V, unsigned PrefAlign,
923 const DataLayout *DL) {
924 assert(V->getType()->isPointerTy() &&
925 "getOrEnforceKnownAlignment expects a pointer!");
926 unsigned BitWidth = DL ? DL->getPointerTypeSizeInBits(V->getType()) : 64;
928 APInt KnownZero(BitWidth, 0), KnownOne(BitWidth, 0);
929 ComputeMaskedBits(V, KnownZero, KnownOne, DL);
930 unsigned TrailZ = KnownZero.countTrailingOnes();
932 // Avoid trouble with ridiculously large TrailZ values, such as
933 // those computed from a null pointer.
934 TrailZ = std::min(TrailZ, unsigned(sizeof(unsigned) * CHAR_BIT - 1));
936 unsigned Align = 1u << std::min(BitWidth - 1, TrailZ);
938 // LLVM doesn't support alignments larger than this currently.
939 Align = std::min(Align, +Value::MaximumAlignment);
941 if (PrefAlign > Align)
942 Align = enforceKnownAlignment(V, Align, PrefAlign, DL);
944 // We don't need to make any adjustment.
945 return Align;
946 }
948 ///===---------------------------------------------------------------------===//
949 /// Dbg Intrinsic utilities
950 ///
952 /// See if there is a dbg.value intrinsic for DIVar before I.
953 static bool LdStHasDebugValue(DIVariable &DIVar, Instruction *I) {
954 // Since we can't guarantee that the original dbg.declare instrinsic
955 // is removed by LowerDbgDeclare(), we need to make sure that we are
956 // not inserting the same dbg.value intrinsic over and over.
957 llvm::BasicBlock::InstListType::iterator PrevI(I);
958 if (PrevI != I->getParent()->getInstList().begin()) {
959 --PrevI;
960 if (DbgValueInst *DVI = dyn_cast<DbgValueInst>(PrevI))
961 if (DVI->getValue() == I->getOperand(0) &&
962 DVI->getOffset() == 0 &&
963 DVI->getVariable() == DIVar)
964 return true;
965 }
966 return false;
967 }
969 /// Inserts a llvm.dbg.value intrinsic before a store to an alloca'd value
970 /// that has an associated llvm.dbg.decl intrinsic.
971 bool llvm::ConvertDebugDeclareToDebugValue(DbgDeclareInst *DDI,
972 StoreInst *SI, DIBuilder &Builder) {
973 DIVariable DIVar(DDI->getVariable());
974 assert((!DIVar || DIVar.isVariable()) &&
975 "Variable in DbgDeclareInst should be either null or a DIVariable.");
976 if (!DIVar)
977 return false;
979 if (LdStHasDebugValue(DIVar, SI))
980 return true;
982 Instruction *DbgVal = NULL;
983 // If an argument is zero extended then use argument directly. The ZExt
984 // may be zapped by an optimization pass in future.
985 Argument *ExtendedArg = NULL;
986 if (ZExtInst *ZExt = dyn_cast<ZExtInst>(SI->getOperand(0)))
987 ExtendedArg = dyn_cast<Argument>(ZExt->getOperand(0));
988 if (SExtInst *SExt = dyn_cast<SExtInst>(SI->getOperand(0)))
989 ExtendedArg = dyn_cast<Argument>(SExt->getOperand(0));
990 if (ExtendedArg)
991 DbgVal = Builder.insertDbgValueIntrinsic(ExtendedArg, 0, DIVar, SI);
992 else
993 DbgVal = Builder.insertDbgValueIntrinsic(SI->getOperand(0), 0, DIVar, SI);
995 // Propagate any debug metadata from the store onto the dbg.value.
996 DebugLoc SIDL = SI->getDebugLoc();
997 if (!SIDL.isUnknown())
998 DbgVal->setDebugLoc(SIDL);
999 // Otherwise propagate debug metadata from dbg.declare.
1000 else
1001 DbgVal->setDebugLoc(DDI->getDebugLoc());
1002 return true;
1003 }
1005 /// Inserts a llvm.dbg.value intrinsic before a load of an alloca'd value
1006 /// that has an associated llvm.dbg.decl intrinsic.
1007 bool llvm::ConvertDebugDeclareToDebugValue(DbgDeclareInst *DDI,
1008 LoadInst *LI, DIBuilder &Builder) {
1009 DIVariable DIVar(DDI->getVariable());
1010 assert((!DIVar || DIVar.isVariable()) &&
1011 "Variable in DbgDeclareInst should be either null or a DIVariable.");
1012 if (!DIVar)
1013 return false;
1015 if (LdStHasDebugValue(DIVar, LI))
1016 return true;
1018 Instruction *DbgVal =
1019 Builder.insertDbgValueIntrinsic(LI->getOperand(0), 0,
1020 DIVar, LI);
1022 // Propagate any debug metadata from the store onto the dbg.value.
1023 DebugLoc LIDL = LI->getDebugLoc();
1024 if (!LIDL.isUnknown())
1025 DbgVal->setDebugLoc(LIDL);
1026 // Otherwise propagate debug metadata from dbg.declare.
1027 else
1028 DbgVal->setDebugLoc(DDI->getDebugLoc());
1029 return true;
1030 }
1032 /// LowerDbgDeclare - Lowers llvm.dbg.declare intrinsics into appropriate set
1033 /// of llvm.dbg.value intrinsics.
1034 bool llvm::LowerDbgDeclare(Function &F) {
1035 DIBuilder DIB(*F.getParent());
1036 SmallVector<DbgDeclareInst *, 4> Dbgs;
1037 for (Function::iterator FI = F.begin(), FE = F.end(); FI != FE; ++FI)
1038 for (BasicBlock::iterator BI = FI->begin(), BE = FI->end(); BI != BE; ++BI) {
1039 if (DbgDeclareInst *DDI = dyn_cast<DbgDeclareInst>(BI))
1040 Dbgs.push_back(DDI);
1041 }
1042 if (Dbgs.empty())
1043 return false;
1045 for (SmallVectorImpl<DbgDeclareInst *>::iterator I = Dbgs.begin(),
1046 E = Dbgs.end(); I != E; ++I) {
1047 DbgDeclareInst *DDI = *I;
1048 if (AllocaInst *AI = dyn_cast_or_null<AllocaInst>(DDI->getAddress())) {
1049 // We only remove the dbg.declare intrinsic if all uses are
1050 // converted to dbg.value intrinsics.
1051 bool RemoveDDI = true;
1052 for (Value::use_iterator UI = AI->use_begin(), E = AI->use_end();
1053 UI != E; ++UI)
1054 if (StoreInst *SI = dyn_cast<StoreInst>(*UI))
1055 ConvertDebugDeclareToDebugValue(DDI, SI, DIB);
1056 else if (LoadInst *LI = dyn_cast<LoadInst>(*UI))
1057 ConvertDebugDeclareToDebugValue(DDI, LI, DIB);
1058 else
1059 RemoveDDI = false;
1060 if (RemoveDDI)
1061 DDI->eraseFromParent();
1062 }
1063 }
1064 return true;
1065 }
1067 /// FindAllocaDbgDeclare - Finds the llvm.dbg.declare intrinsic describing the
1068 /// alloca 'V', if any.
1069 DbgDeclareInst *llvm::FindAllocaDbgDeclare(Value *V) {
1070 if (MDNode *DebugNode = MDNode::getIfExists(V->getContext(), V))
1071 for (Value::use_iterator UI = DebugNode->use_begin(),
1072 E = DebugNode->use_end(); UI != E; ++UI)
1073 if (DbgDeclareInst *DDI = dyn_cast<DbgDeclareInst>(*UI))
1074 return DDI;
1076 return 0;
1077 }
1079 bool llvm::replaceDbgDeclareForAlloca(AllocaInst *AI, Value *NewAllocaAddress,
1080 DIBuilder &Builder) {
1081 DbgDeclareInst *DDI = FindAllocaDbgDeclare(AI);
1082 if (!DDI)
1083 return false;
1084 DIVariable DIVar(DDI->getVariable());
1085 assert((!DIVar || DIVar.isVariable()) &&
1086 "Variable in DbgDeclareInst should be either null or a DIVariable.");
1087 if (!DIVar)
1088 return false;
1090 // Create a copy of the original DIDescriptor for user variable, appending
1091 // "deref" operation to a list of address elements, as new llvm.dbg.declare
1092 // will take a value storing address of the memory for variable, not
1093 // alloca itself.
1094 Type *Int64Ty = Type::getInt64Ty(AI->getContext());
1095 SmallVector<Value*, 4> NewDIVarAddress;
1096 if (DIVar.hasComplexAddress()) {
1097 for (unsigned i = 0, n = DIVar.getNumAddrElements(); i < n; ++i) {
1098 NewDIVarAddress.push_back(
1099 ConstantInt::get(Int64Ty, DIVar.getAddrElement(i)));
1100 }
1101 }
1102 NewDIVarAddress.push_back(ConstantInt::get(Int64Ty, DIBuilder::OpDeref));
1103 DIVariable NewDIVar = Builder.createComplexVariable(
1104 DIVar.getTag(), DIVar.getContext(), DIVar.getName(),
1105 DIVar.getFile(), DIVar.getLineNumber(), DIVar.getType(),
1106 NewDIVarAddress, DIVar.getArgNumber());
1108 // Insert llvm.dbg.declare in the same basic block as the original alloca,
1109 // and remove old llvm.dbg.declare.
1110 BasicBlock *BB = AI->getParent();
1111 Builder.insertDeclare(NewAllocaAddress, NewDIVar, BB);
1112 DDI->eraseFromParent();
1113 return true;
1114 }
1116 /// changeToUnreachable - Insert an unreachable instruction before the specified
1117 /// instruction, making it and the rest of the code in the block dead.
1118 static void changeToUnreachable(Instruction *I, bool UseLLVMTrap) {
1119 BasicBlock *BB = I->getParent();
1120 // Loop over all of the successors, removing BB's entry from any PHI
1121 // nodes.
1122 for (succ_iterator SI = succ_begin(BB), SE = succ_end(BB); SI != SE; ++SI)
1123 (*SI)->removePredecessor(BB);
1125 // Insert a call to llvm.trap right before this. This turns the undefined
1126 // behavior into a hard fail instead of falling through into random code.
1127 if (UseLLVMTrap) {
1128 Function *TrapFn =
1129 Intrinsic::getDeclaration(BB->getParent()->getParent(), Intrinsic::trap);
1130 CallInst *CallTrap = CallInst::Create(TrapFn, "", I);
1131 CallTrap->setDebugLoc(I->getDebugLoc());
1132 }
1133 new UnreachableInst(I->getContext(), I);
1135 // All instructions after this are dead.
1136 BasicBlock::iterator BBI = I, BBE = BB->end();
1137 while (BBI != BBE) {
1138 if (!BBI->use_empty())
1139 BBI->replaceAllUsesWith(UndefValue::get(BBI->getType()));
1140 BB->getInstList().erase(BBI++);
1141 }
1142 }
1144 /// changeToCall - Convert the specified invoke into a normal call.
1145 static void changeToCall(InvokeInst *II) {
1146 SmallVector<Value*, 8> Args(II->op_begin(), II->op_end() - 3);
1147 CallInst *NewCall = CallInst::Create(II->getCalledValue(), Args, "", II);
1148 NewCall->takeName(II);
1149 NewCall->setCallingConv(II->getCallingConv());
1150 NewCall->setAttributes(II->getAttributes());
1151 NewCall->setDebugLoc(II->getDebugLoc());
1152 II->replaceAllUsesWith(NewCall);
1154 // Follow the call by a branch to the normal destination.
1155 BranchInst::Create(II->getNormalDest(), II);
1157 // Update PHI nodes in the unwind destination
1158 II->getUnwindDest()->removePredecessor(II->getParent());
1159 II->eraseFromParent();
1160 }
1162 static bool markAliveBlocks(BasicBlock *BB,
1163 SmallPtrSet<BasicBlock*, 128> &Reachable) {
1165 SmallVector<BasicBlock*, 128> Worklist;
1166 Worklist.push_back(BB);
1167 Reachable.insert(BB);
1168 bool Changed = false;
1169 do {
1170 BB = Worklist.pop_back_val();
1172 // Do a quick scan of the basic block, turning any obviously unreachable
1173 // instructions into LLVM unreachable insts. The instruction combining pass
1174 // canonicalizes unreachable insts into stores to null or undef.
1175 for (BasicBlock::iterator BBI = BB->begin(), E = BB->end(); BBI != E;++BBI){
1176 if (CallInst *CI = dyn_cast<CallInst>(BBI)) {
1177 if (CI->doesNotReturn()) {
1178 // If we found a call to a no-return function, insert an unreachable
1179 // instruction after it. Make sure there isn't *already* one there
1180 // though.
1181 ++BBI;
1182 if (!isa<UnreachableInst>(BBI)) {
1183 // Don't insert a call to llvm.trap right before the unreachable.
1184 changeToUnreachable(BBI, false);
1185 Changed = true;
1186 }
1187 break;
1188 }
1189 }
1191 // Store to undef and store to null are undefined and used to signal that
1192 // they should be changed to unreachable by passes that can't modify the
1193 // CFG.
1194 if (StoreInst *SI = dyn_cast<StoreInst>(BBI)) {
1195 // Don't touch volatile stores.
1196 if (SI->isVolatile()) continue;
1198 Value *Ptr = SI->getOperand(1);
1200 if (isa<UndefValue>(Ptr) ||
1201 (isa<ConstantPointerNull>(Ptr) &&
1202 SI->getPointerAddressSpace() == 0)) {
1203 changeToUnreachable(SI, true);
1204 Changed = true;
1205 break;
1206 }
1207 }
1208 }
1210 // Turn invokes that call 'nounwind' functions into ordinary calls.
1211 if (InvokeInst *II = dyn_cast<InvokeInst>(BB->getTerminator())) {
1212 Value *Callee = II->getCalledValue();
1213 if (isa<ConstantPointerNull>(Callee) || isa<UndefValue>(Callee)) {
1214 changeToUnreachable(II, true);
1215 Changed = true;
1216 } else if (II->doesNotThrow()) {
1217 if (II->use_empty() && II->onlyReadsMemory()) {
1218 // jump to the normal destination branch.
1219 BranchInst::Create(II->getNormalDest(), II);
1220 II->getUnwindDest()->removePredecessor(II->getParent());
1221 II->eraseFromParent();
1222 } else
1223 changeToCall(II);
1224 Changed = true;
1225 }
1226 }
1228 Changed |= ConstantFoldTerminator(BB, true);
1229 for (succ_iterator SI = succ_begin(BB), SE = succ_end(BB); SI != SE; ++SI)
1230 if (Reachable.insert(*SI))
1231 Worklist.push_back(*SI);
1232 } while (!Worklist.empty());
1233 return Changed;
1234 }
1236 /// removeUnreachableBlocksFromFn - Remove blocks that are not reachable, even
1237 /// if they are in a dead cycle. Return true if a change was made, false
1238 /// otherwise.
1239 bool llvm::removeUnreachableBlocks(Function &F) {
1240 SmallPtrSet<BasicBlock*, 128> Reachable;
1241 bool Changed = markAliveBlocks(F.begin(), Reachable);
1243 // If there are unreachable blocks in the CFG...
1244 if (Reachable.size() == F.size())
1245 return Changed;
1247 assert(Reachable.size() < F.size());
1248 NumRemoved += F.size()-Reachable.size();
1250 // Loop over all of the basic blocks that are not reachable, dropping all of
1251 // their internal references...
1252 for (Function::iterator BB = ++F.begin(), E = F.end(); BB != E; ++BB) {
1253 if (Reachable.count(BB))
1254 continue;
1256 for (succ_iterator SI = succ_begin(BB), SE = succ_end(BB); SI != SE; ++SI)
1257 if (Reachable.count(*SI))
1258 (*SI)->removePredecessor(BB);
1259 BB->dropAllReferences();
1260 }
1262 for (Function::iterator I = ++F.begin(); I != F.end();)
1263 if (!Reachable.count(I))
1264 I = F.getBasicBlockList().erase(I);
1265 else
1266 ++I;
1268 return true;
1269 }