1 //===- InstCombineLoadStoreAlloca.cpp -------------------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements the visit functions for load, store and alloca.
12 //===----------------------------------------------------------------------===//
14 #include "InstCombine.h"
15 #include "llvm/IntrinsicInst.h"
16 #include "llvm/Analysis/Loads.h"
17 #include "llvm/Target/TargetData.h"
18 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
19 #include "llvm/Transforms/Utils/Local.h"
20 #include "llvm/ADT/Statistic.h"
23 STATISTIC(NumDeadStore, "Number of dead stores eliminated");
24 STATISTIC(NumGlobalCopies, "Number of allocas copied from constant global");
26 /// pointsToConstantGlobal - Return true if V (possibly indirectly) points to
27 /// some part of a constant global variable. This intentionally only accepts
28 /// constant expressions because we can't rewrite arbitrary instructions.
29 static bool pointsToConstantGlobal(Value *V) {
30 if (GlobalVariable *GV = dyn_cast<GlobalVariable>(V))
31 return GV->isConstant();
32 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V))
33 if (CE->getOpcode() == Instruction::BitCast ||
34 CE->getOpcode() == Instruction::GetElementPtr)
35 return pointsToConstantGlobal(CE->getOperand(0));
39 /// isOnlyCopiedFromConstantGlobal - Recursively walk the uses of a (derived)
40 /// pointer to an alloca. Ignore any reads of the pointer, return false if we
41 /// see any stores or other unknown uses. If we see pointer arithmetic, keep
42 /// track of whether it moves the pointer (with IsOffset) but otherwise traverse
43 /// the uses. If we see a memcpy/memmove that targets an unoffseted pointer to
44 /// the alloca, and if the source pointer is a pointer to a constant global, we
45 /// can optimize this.
47 isOnlyCopiedFromConstantGlobal(Value *V, MemTransferInst *&TheCopy,
48 SmallVectorImpl<Instruction *> &ToDelete,
49 bool IsOffset = false) {
50 // We track lifetime intrinsics as we encounter them. If we decide to go
51 // ahead and replace the value with the global, this lets the caller quickly
52 // eliminate the markers.
54 for (Value::use_iterator UI = V->use_begin(), E = V->use_end(); UI!=E; ++UI) {
55 User *U = cast<Instruction>(*UI);
57 if (LoadInst *LI = dyn_cast<LoadInst>(U)) {
58 // Ignore non-volatile loads, they are always ok.
59 if (!LI->isSimple()) return false;
63 if (BitCastInst *BCI = dyn_cast<BitCastInst>(U)) {
64 // If uses of the bitcast are ok, we are ok.
65 if (!isOnlyCopiedFromConstantGlobal(BCI, TheCopy, ToDelete, IsOffset))
69 if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(U)) {
70 // If the GEP has all zero indices, it doesn't offset the pointer. If it
72 if (!isOnlyCopiedFromConstantGlobal(GEP, TheCopy, ToDelete,
73 IsOffset || !GEP->hasAllZeroIndices()))
78 if (CallSite CS = U) {
79 // If this is the function being called then we treat it like a load and
84 // If this is a readonly/readnone call site, then we know it is just a
85 // load (but one that potentially returns the value itself), so we can
86 // ignore it if we know that the value isn't captured.
87 unsigned ArgNo = CS.getArgumentNo(UI);
88 if (CS.onlyReadsMemory() &&
89 (CS.getInstruction()->use_empty() || CS.doesNotCapture(ArgNo)))
92 // If this is being passed as a byval argument, the caller is making a
93 // copy, so it is only a read of the alloca.
94 if (CS.isByValArgument(ArgNo))
98 // Lifetime intrinsics can be handled by the caller.
99 if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(U)) {
100 if (II->getIntrinsicID() == Intrinsic::lifetime_start ||
101 II->getIntrinsicID() == Intrinsic::lifetime_end) {
102 assert(II->use_empty() && "Lifetime markers have no result to use!");
103 ToDelete.push_back(II);
108 // If this is isn't our memcpy/memmove, reject it as something we can't
110 MemTransferInst *MI = dyn_cast<MemTransferInst>(U);
114 // If the transfer is using the alloca as a source of the transfer, then
115 // ignore it since it is a load (unless the transfer is volatile).
116 if (UI.getOperandNo() == 1) {
117 if (MI->isVolatile()) return false;
121 // If we already have seen a copy, reject the second one.
122 if (TheCopy) return false;
124 // If the pointer has been offset from the start of the alloca, we can't
125 // safely handle this.
126 if (IsOffset) return false;
128 // If the memintrinsic isn't using the alloca as the dest, reject it.
129 if (UI.getOperandNo() != 0) return false;
131 // If the source of the memcpy/move is not a constant global, reject it.
132 if (!pointsToConstantGlobal(MI->getSource()))
135 // Otherwise, the transform is safe. Remember the copy instruction.
141 /// isOnlyCopiedFromConstantGlobal - Return true if the specified alloca is only
142 /// modified by a copy from a constant global. If we can prove this, we can
143 /// replace any uses of the alloca with uses of the global directly.
144 static MemTransferInst *
145 isOnlyCopiedFromConstantGlobal(AllocaInst *AI,
146 SmallVectorImpl<Instruction *> &ToDelete) {
147 MemTransferInst *TheCopy = 0;
148 if (isOnlyCopiedFromConstantGlobal(AI, TheCopy, ToDelete))
153 /// getPointeeAlignment - Compute the minimum alignment of the value pointed
154 /// to by the given pointer.
155 static unsigned getPointeeAlignment(Value *V, const TargetData &TD) {
156 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V))
157 if (CE->getOpcode() == Instruction::BitCast ||
158 (CE->getOpcode() == Instruction::GetElementPtr &&
159 cast<GEPOperator>(CE)->hasAllZeroIndices()))
160 return getPointeeAlignment(CE->getOperand(0), TD);
162 if (GlobalVariable *GV = dyn_cast<GlobalVariable>(V))
163 if (!GV->isDeclaration())
164 return TD.getPreferredAlignment(GV);
166 if (PointerType *PT = dyn_cast<PointerType>(V->getType()))
167 return TD.getABITypeAlignment(PT->getElementType());
172 Instruction *InstCombiner::visitAllocaInst(AllocaInst &AI) {
173 // Ensure that the alloca array size argument has type intptr_t, so that
174 // any casting is exposed early.
176 Type *IntPtrTy = TD->getIntPtrType(AI.getContext());
177 if (AI.getArraySize()->getType() != IntPtrTy) {
178 Value *V = Builder->CreateIntCast(AI.getArraySize(),
185 // Convert: alloca Ty, C - where C is a constant != 1 into: alloca [C x Ty], 1
186 if (AI.isArrayAllocation()) { // Check C != 1
187 if (const ConstantInt *C = dyn_cast<ConstantInt>(AI.getArraySize())) {
189 ArrayType::get(AI.getAllocatedType(), C->getZExtValue());
190 AllocaInst *New = Builder->CreateAlloca(NewTy, 0, AI.getName());
191 New->setAlignment(AI.getAlignment());
193 // Scan to the end of the allocation instructions, to skip over a block of
194 // allocas if possible...also skip interleaved debug info
196 BasicBlock::iterator It = New;
197 while (isa<AllocaInst>(*It) || isa<DbgInfoIntrinsic>(*It)) ++It;
199 // Now that I is pointing to the first non-allocation-inst in the block,
200 // insert our getelementptr instruction...
202 Value *NullIdx =Constant::getNullValue(Type::getInt32Ty(AI.getContext()));
207 GetElementPtrInst::CreateInBounds(New, Idx, New->getName()+".sub");
208 InsertNewInstBefore(GEP, *It);
210 // Now make everything use the getelementptr instead of the original
212 return ReplaceInstUsesWith(AI, GEP);
213 } else if (isa<UndefValue>(AI.getArraySize())) {
214 return ReplaceInstUsesWith(AI, Constant::getNullValue(AI.getType()));
218 if (TD && AI.getAllocatedType()->isSized()) {
219 // If the alignment is 0 (unspecified), assign it the preferred alignment.
220 if (AI.getAlignment() == 0)
221 AI.setAlignment(TD->getPrefTypeAlignment(AI.getAllocatedType()));
223 // Move all alloca's of zero byte objects to the entry block and merge them
224 // together. Note that we only do this for alloca's, because malloc should
225 // allocate and return a unique pointer, even for a zero byte allocation.
226 if (TD->getTypeAllocSize(AI.getAllocatedType()) == 0) {
227 // For a zero sized alloca there is no point in doing an array allocation.
228 // This is helpful if the array size is a complicated expression not used
230 if (AI.isArrayAllocation()) {
231 AI.setOperand(0, ConstantInt::get(AI.getArraySize()->getType(), 1));
235 // Get the first instruction in the entry block.
236 BasicBlock &EntryBlock = AI.getParent()->getParent()->getEntryBlock();
237 Instruction *FirstInst = EntryBlock.getFirstNonPHIOrDbg();
238 if (FirstInst != &AI) {
239 // If the entry block doesn't start with a zero-size alloca then move
240 // this one to the start of the entry block. There is no problem with
241 // dominance as the array size was forced to a constant earlier already.
242 AllocaInst *EntryAI = dyn_cast<AllocaInst>(FirstInst);
243 if (!EntryAI || !EntryAI->getAllocatedType()->isSized() ||
244 TD->getTypeAllocSize(EntryAI->getAllocatedType()) != 0) {
245 AI.moveBefore(FirstInst);
249 // If the alignment of the entry block alloca is 0 (unspecified),
250 // assign it the preferred alignment.
251 if (EntryAI->getAlignment() == 0)
252 EntryAI->setAlignment(
253 TD->getPrefTypeAlignment(EntryAI->getAllocatedType()));
254 // Replace this zero-sized alloca with the one at the start of the entry
255 // block after ensuring that the address will be aligned enough for both
257 unsigned MaxAlign = std::max(EntryAI->getAlignment(),
259 EntryAI->setAlignment(MaxAlign);
260 if (AI.getType() != EntryAI->getType())
261 return new BitCastInst(EntryAI, AI.getType());
262 return ReplaceInstUsesWith(AI, EntryAI);
268 // Check to see if this allocation is only modified by a memcpy/memmove from
269 // a constant global whose alignment is equal to or exceeds that of the
270 // allocation. If this is the case, we can change all users to use
271 // the constant global instead. This is commonly produced by the CFE by
272 // constructs like "void foo() { int A[] = {1,2,3,4,5,6,7,8,9...}; }" if 'A'
273 // is only subsequently read.
274 SmallVector<Instruction *, 4> ToDelete;
275 if (MemTransferInst *Copy = isOnlyCopiedFromConstantGlobal(&AI, ToDelete)) {
276 if (AI.getAlignment() <= getPointeeAlignment(Copy->getSource(), *TD)) {
277 DEBUG(dbgs() << "Found alloca equal to global: " << AI << '\n');
278 DEBUG(dbgs() << " memcpy = " << *Copy << '\n');
279 for (unsigned i = 0, e = ToDelete.size(); i != e; ++i)
280 EraseInstFromFunction(*ToDelete[i]);
281 Constant *TheSrc = cast<Constant>(Copy->getSource());
283 = ReplaceInstUsesWith(AI, ConstantExpr::getBitCast(TheSrc,
285 EraseInstFromFunction(*Copy);
292 // At last, use the generic allocation site handler to aggressively remove
294 return visitAllocSite(AI);
298 /// InstCombineLoadCast - Fold 'load (cast P)' -> cast (load P)' when possible.
299 static Instruction *InstCombineLoadCast(InstCombiner &IC, LoadInst &LI,
300 const TargetData *TD) {
301 User *CI = cast<User>(LI.getOperand(0));
302 Value *CastOp = CI->getOperand(0);
304 PointerType *DestTy = cast<PointerType>(CI->getType());
305 Type *DestPTy = DestTy->getElementType();
306 if (PointerType *SrcTy = dyn_cast<PointerType>(CastOp->getType())) {
308 // If the address spaces don't match, don't eliminate the cast.
309 if (DestTy->getAddressSpace() != SrcTy->getAddressSpace())
312 Type *SrcPTy = SrcTy->getElementType();
314 if (DestPTy->isIntegerTy() || DestPTy->isPointerTy() ||
315 DestPTy->isVectorTy()) {
316 // If the source is an array, the code below will not succeed. Check to
317 // see if a trivial 'gep P, 0, 0' will help matters. Only do this for
319 if (ArrayType *ASrcTy = dyn_cast<ArrayType>(SrcPTy))
320 if (Constant *CSrc = dyn_cast<Constant>(CastOp))
321 if (ASrcTy->getNumElements() != 0) {
323 Idxs[0] = Constant::getNullValue(Type::getInt32Ty(LI.getContext()));
325 CastOp = ConstantExpr::getGetElementPtr(CSrc, Idxs);
326 SrcTy = cast<PointerType>(CastOp->getType());
327 SrcPTy = SrcTy->getElementType();
330 if (IC.getTargetData() &&
331 (SrcPTy->isIntegerTy() || SrcPTy->isPointerTy() ||
332 SrcPTy->isVectorTy()) &&
333 // Do not allow turning this into a load of an integer, which is then
334 // casted to a pointer, this pessimizes pointer analysis a lot.
335 (SrcPTy->isPointerTy() == LI.getType()->isPointerTy()) &&
336 IC.getTargetData()->getTypeSizeInBits(SrcPTy) ==
337 IC.getTargetData()->getTypeSizeInBits(DestPTy)) {
339 // Okay, we are casting from one integer or pointer type to another of
340 // the same size. Instead of casting the pointer before the load, cast
341 // the result of the loaded value.
343 IC.Builder->CreateLoad(CastOp, LI.isVolatile(), CI->getName());
344 NewLoad->setAlignment(LI.getAlignment());
345 NewLoad->setAtomic(LI.getOrdering(), LI.getSynchScope());
346 // Now cast the result of the load.
347 return new BitCastInst(NewLoad, LI.getType());
354 Instruction *InstCombiner::visitLoadInst(LoadInst &LI) {
355 Value *Op = LI.getOperand(0);
357 // Attempt to improve the alignment.
359 unsigned KnownAlign =
360 getOrEnforceKnownAlignment(Op, TD->getPrefTypeAlignment(LI.getType()),TD);
361 unsigned LoadAlign = LI.getAlignment();
362 unsigned EffectiveLoadAlign = LoadAlign != 0 ? LoadAlign :
363 TD->getABITypeAlignment(LI.getType());
365 if (KnownAlign > EffectiveLoadAlign)
366 LI.setAlignment(KnownAlign);
367 else if (LoadAlign == 0)
368 LI.setAlignment(EffectiveLoadAlign);
371 // load (cast X) --> cast (load X) if safe.
372 if (isa<CastInst>(Op))
373 if (Instruction *Res = InstCombineLoadCast(*this, LI, TD))
376 // None of the following transforms are legal for volatile/atomic loads.
377 // FIXME: Some of it is okay for atomic loads; needs refactoring.
378 if (!LI.isSimple()) return 0;
380 // Do really simple store-to-load forwarding and load CSE, to catch cases
381 // where there are several consecutive memory accesses to the same location,
382 // separated by a few arithmetic operations.
383 BasicBlock::iterator BBI = &LI;
384 if (Value *AvailableVal = FindAvailableLoadedValue(Op, LI.getParent(), BBI,6))
385 return ReplaceInstUsesWith(LI, AvailableVal);
387 // load(gep null, ...) -> unreachable
388 if (GetElementPtrInst *GEPI = dyn_cast<GetElementPtrInst>(Op)) {
389 const Value *GEPI0 = GEPI->getOperand(0);
390 // TODO: Consider a target hook for valid address spaces for this xform.
391 if (isa<ConstantPointerNull>(GEPI0) && GEPI->getPointerAddressSpace() == 0){
392 // Insert a new store to null instruction before the load to indicate
393 // that this code is not reachable. We do this instead of inserting
394 // an unreachable instruction directly because we cannot modify the
396 new StoreInst(UndefValue::get(LI.getType()),
397 Constant::getNullValue(Op->getType()), &LI);
398 return ReplaceInstUsesWith(LI, UndefValue::get(LI.getType()));
402 // load null/undef -> unreachable
403 // TODO: Consider a target hook for valid address spaces for this xform.
404 if (isa<UndefValue>(Op) ||
405 (isa<ConstantPointerNull>(Op) && LI.getPointerAddressSpace() == 0)) {
406 // Insert a new store to null instruction before the load to indicate that
407 // this code is not reachable. We do this instead of inserting an
408 // unreachable instruction directly because we cannot modify the CFG.
409 new StoreInst(UndefValue::get(LI.getType()),
410 Constant::getNullValue(Op->getType()), &LI);
411 return ReplaceInstUsesWith(LI, UndefValue::get(LI.getType()));
414 // Instcombine load (constantexpr_cast global) -> cast (load global)
415 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Op))
417 if (Instruction *Res = InstCombineLoadCast(*this, LI, TD))
420 if (Op->hasOneUse()) {
421 // Change select and PHI nodes to select values instead of addresses: this
422 // helps alias analysis out a lot, allows many others simplifications, and
423 // exposes redundancy in the code.
425 // Note that we cannot do the transformation unless we know that the
426 // introduced loads cannot trap! Something like this is valid as long as
427 // the condition is always false: load (select bool %C, int* null, int* %G),
428 // but it would not be valid if we transformed it to load from null
431 if (SelectInst *SI = dyn_cast<SelectInst>(Op)) {
432 // load (select (Cond, &V1, &V2)) --> select(Cond, load &V1, load &V2).
433 unsigned Align = LI.getAlignment();
434 if (isSafeToLoadUnconditionally(SI->getOperand(1), SI, Align, TD) &&
435 isSafeToLoadUnconditionally(SI->getOperand(2), SI, Align, TD)) {
436 LoadInst *V1 = Builder->CreateLoad(SI->getOperand(1),
437 SI->getOperand(1)->getName()+".val");
438 LoadInst *V2 = Builder->CreateLoad(SI->getOperand(2),
439 SI->getOperand(2)->getName()+".val");
440 V1->setAlignment(Align);
441 V2->setAlignment(Align);
442 return SelectInst::Create(SI->getCondition(), V1, V2);
445 // load (select (cond, null, P)) -> load P
446 if (Constant *C = dyn_cast<Constant>(SI->getOperand(1)))
447 if (C->isNullValue()) {
448 LI.setOperand(0, SI->getOperand(2));
452 // load (select (cond, P, null)) -> load P
453 if (Constant *C = dyn_cast<Constant>(SI->getOperand(2)))
454 if (C->isNullValue()) {
455 LI.setOperand(0, SI->getOperand(1));
463 /// InstCombineStoreToCast - Fold store V, (cast P) -> store (cast V), P
464 /// when possible. This makes it generally easy to do alias analysis and/or
465 /// SROA/mem2reg of the memory object.
466 static Instruction *InstCombineStoreToCast(InstCombiner &IC, StoreInst &SI) {
467 User *CI = cast<User>(SI.getOperand(1));
468 Value *CastOp = CI->getOperand(0);
470 Type *DestPTy = cast<PointerType>(CI->getType())->getElementType();
471 PointerType *SrcTy = dyn_cast<PointerType>(CastOp->getType());
472 if (SrcTy == 0) return 0;
474 Type *SrcPTy = SrcTy->getElementType();
476 if (!DestPTy->isIntegerTy() && !DestPTy->isPointerTy())
479 /// NewGEPIndices - If SrcPTy is an aggregate type, we can emit a "noop gep"
480 /// to its first element. This allows us to handle things like:
481 /// store i32 xxx, (bitcast {foo*, float}* %P to i32*)
483 SmallVector<Value*, 4> NewGEPIndices;
485 // If the source is an array, the code below will not succeed. Check to
486 // see if a trivial 'gep P, 0, 0' will help matters. Only do this for
488 if (SrcPTy->isArrayTy() || SrcPTy->isStructTy()) {
489 // Index through pointer.
490 Constant *Zero = Constant::getNullValue(Type::getInt32Ty(SI.getContext()));
491 NewGEPIndices.push_back(Zero);
494 if (StructType *STy = dyn_cast<StructType>(SrcPTy)) {
495 if (!STy->getNumElements()) /* Struct can be empty {} */
497 NewGEPIndices.push_back(Zero);
498 SrcPTy = STy->getElementType(0);
499 } else if (ArrayType *ATy = dyn_cast<ArrayType>(SrcPTy)) {
500 NewGEPIndices.push_back(Zero);
501 SrcPTy = ATy->getElementType();
507 SrcTy = PointerType::get(SrcPTy, SrcTy->getAddressSpace());
510 if (!SrcPTy->isIntegerTy() && !SrcPTy->isPointerTy())
513 // If the pointers point into different address spaces or if they point to
514 // values with different sizes, we can't do the transformation.
515 if (!IC.getTargetData() ||
516 SrcTy->getAddressSpace() !=
517 cast<PointerType>(CI->getType())->getAddressSpace() ||
518 IC.getTargetData()->getTypeSizeInBits(SrcPTy) !=
519 IC.getTargetData()->getTypeSizeInBits(DestPTy))
522 // Okay, we are casting from one integer or pointer type to another of
523 // the same size. Instead of casting the pointer before
524 // the store, cast the value to be stored.
526 Value *SIOp0 = SI.getOperand(0);
527 Instruction::CastOps opcode = Instruction::BitCast;
528 Type* CastSrcTy = SIOp0->getType();
529 Type* CastDstTy = SrcPTy;
530 if (CastDstTy->isPointerTy()) {
531 if (CastSrcTy->isIntegerTy())
532 opcode = Instruction::IntToPtr;
533 } else if (CastDstTy->isIntegerTy()) {
534 if (SIOp0->getType()->isPointerTy())
535 opcode = Instruction::PtrToInt;
538 // SIOp0 is a pointer to aggregate and this is a store to the first field,
539 // emit a GEP to index into its first field.
540 if (!NewGEPIndices.empty())
541 CastOp = IC.Builder->CreateInBoundsGEP(CastOp, NewGEPIndices);
543 NewCast = IC.Builder->CreateCast(opcode, SIOp0, CastDstTy,
544 SIOp0->getName()+".c");
545 SI.setOperand(0, NewCast);
546 SI.setOperand(1, CastOp);
550 /// equivalentAddressValues - Test if A and B will obviously have the same
551 /// value. This includes recognizing that %t0 and %t1 will have the same
552 /// value in code like this:
553 /// %t0 = getelementptr \@a, 0, 3
554 /// store i32 0, i32* %t0
555 /// %t1 = getelementptr \@a, 0, 3
556 /// %t2 = load i32* %t1
558 static bool equivalentAddressValues(Value *A, Value *B) {
559 // Test if the values are trivially equivalent.
560 if (A == B) return true;
562 // Test if the values come form identical arithmetic instructions.
563 // This uses isIdenticalToWhenDefined instead of isIdenticalTo because
564 // its only used to compare two uses within the same basic block, which
565 // means that they'll always either have the same value or one of them
566 // will have an undefined value.
567 if (isa<BinaryOperator>(A) ||
570 isa<GetElementPtrInst>(A))
571 if (Instruction *BI = dyn_cast<Instruction>(B))
572 if (cast<Instruction>(A)->isIdenticalToWhenDefined(BI))
575 // Otherwise they may not be equivalent.
579 Instruction *InstCombiner::visitStoreInst(StoreInst &SI) {
580 Value *Val = SI.getOperand(0);
581 Value *Ptr = SI.getOperand(1);
583 // Attempt to improve the alignment.
585 unsigned KnownAlign =
586 getOrEnforceKnownAlignment(Ptr, TD->getPrefTypeAlignment(Val->getType()),
588 unsigned StoreAlign = SI.getAlignment();
589 unsigned EffectiveStoreAlign = StoreAlign != 0 ? StoreAlign :
590 TD->getABITypeAlignment(Val->getType());
592 if (KnownAlign > EffectiveStoreAlign)
593 SI.setAlignment(KnownAlign);
594 else if (StoreAlign == 0)
595 SI.setAlignment(EffectiveStoreAlign);
598 // Don't hack volatile/atomic stores.
599 // FIXME: Some bits are legal for atomic stores; needs refactoring.
600 if (!SI.isSimple()) return 0;
602 // If the RHS is an alloca with a single use, zapify the store, making the
604 if (Ptr->hasOneUse()) {
605 if (isa<AllocaInst>(Ptr))
606 return EraseInstFromFunction(SI);
607 if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Ptr)) {
608 if (isa<AllocaInst>(GEP->getOperand(0))) {
609 if (GEP->getOperand(0)->hasOneUse())
610 return EraseInstFromFunction(SI);
615 // Do really simple DSE, to catch cases where there are several consecutive
616 // stores to the same location, separated by a few arithmetic operations. This
617 // situation often occurs with bitfield accesses.
618 BasicBlock::iterator BBI = &SI;
619 for (unsigned ScanInsts = 6; BBI != SI.getParent()->begin() && ScanInsts;
622 // Don't count debug info directives, lest they affect codegen,
623 // and we skip pointer-to-pointer bitcasts, which are NOPs.
624 if (isa<DbgInfoIntrinsic>(BBI) ||
625 (isa<BitCastInst>(BBI) && BBI->getType()->isPointerTy())) {
630 if (StoreInst *PrevSI = dyn_cast<StoreInst>(BBI)) {
631 // Prev store isn't volatile, and stores to the same location?
632 if (PrevSI->isSimple() && equivalentAddressValues(PrevSI->getOperand(1),
636 EraseInstFromFunction(*PrevSI);
642 // If this is a load, we have to stop. However, if the loaded value is from
643 // the pointer we're loading and is producing the pointer we're storing,
644 // then *this* store is dead (X = load P; store X -> P).
645 if (LoadInst *LI = dyn_cast<LoadInst>(BBI)) {
646 if (LI == Val && equivalentAddressValues(LI->getOperand(0), Ptr) &&
648 return EraseInstFromFunction(SI);
650 // Otherwise, this is a load from some other location. Stores before it
655 // Don't skip over loads or things that can modify memory.
656 if (BBI->mayWriteToMemory() || BBI->mayReadFromMemory())
660 // store X, null -> turns into 'unreachable' in SimplifyCFG
661 if (isa<ConstantPointerNull>(Ptr) && SI.getPointerAddressSpace() == 0) {
662 if (!isa<UndefValue>(Val)) {
663 SI.setOperand(0, UndefValue::get(Val->getType()));
664 if (Instruction *U = dyn_cast<Instruction>(Val))
665 Worklist.Add(U); // Dropped a use.
667 return 0; // Do not modify these!
670 // store undef, Ptr -> noop
671 if (isa<UndefValue>(Val))
672 return EraseInstFromFunction(SI);
674 // If the pointer destination is a cast, see if we can fold the cast into the
676 if (isa<CastInst>(Ptr))
677 if (Instruction *Res = InstCombineStoreToCast(*this, SI))
679 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Ptr))
681 if (Instruction *Res = InstCombineStoreToCast(*this, SI))
685 // If this store is the last instruction in the basic block (possibly
686 // excepting debug info instructions), and if the block ends with an
687 // unconditional branch, try to move it to the successor block.
691 } while (isa<DbgInfoIntrinsic>(BBI) ||
692 (isa<BitCastInst>(BBI) && BBI->getType()->isPointerTy()));
693 if (BranchInst *BI = dyn_cast<BranchInst>(BBI))
694 if (BI->isUnconditional())
695 if (SimplifyStoreAtEndOfBlock(SI))
696 return 0; // xform done!
701 /// SimplifyStoreAtEndOfBlock - Turn things like:
702 /// if () { *P = v1; } else { *P = v2 }
703 /// into a phi node with a store in the successor.
705 /// Simplify things like:
706 /// *P = v1; if () { *P = v2; }
707 /// into a phi node with a store in the successor.
709 bool InstCombiner::SimplifyStoreAtEndOfBlock(StoreInst &SI) {
710 BasicBlock *StoreBB = SI.getParent();
712 // Check to see if the successor block has exactly two incoming edges. If
713 // so, see if the other predecessor contains a store to the same location.
714 // if so, insert a PHI node (if needed) and move the stores down.
715 BasicBlock *DestBB = StoreBB->getTerminator()->getSuccessor(0);
717 // Determine whether Dest has exactly two predecessors and, if so, compute
718 // the other predecessor.
719 pred_iterator PI = pred_begin(DestBB);
721 BasicBlock *OtherBB = 0;
726 if (++PI == pred_end(DestBB))
735 if (++PI != pred_end(DestBB))
738 // Bail out if all the relevant blocks aren't distinct (this can happen,
739 // for example, if SI is in an infinite loop)
740 if (StoreBB == DestBB || OtherBB == DestBB)
743 // Verify that the other block ends in a branch and is not otherwise empty.
744 BasicBlock::iterator BBI = OtherBB->getTerminator();
745 BranchInst *OtherBr = dyn_cast<BranchInst>(BBI);
746 if (!OtherBr || BBI == OtherBB->begin())
749 // If the other block ends in an unconditional branch, check for the 'if then
750 // else' case. there is an instruction before the branch.
751 StoreInst *OtherStore = 0;
752 if (OtherBr->isUnconditional()) {
754 // Skip over debugging info.
755 while (isa<DbgInfoIntrinsic>(BBI) ||
756 (isa<BitCastInst>(BBI) && BBI->getType()->isPointerTy())) {
757 if (BBI==OtherBB->begin())
761 // If this isn't a store, isn't a store to the same location, or is not the
762 // right kind of store, bail out.
763 OtherStore = dyn_cast<StoreInst>(BBI);
764 if (!OtherStore || OtherStore->getOperand(1) != SI.getOperand(1) ||
765 !SI.isSameOperationAs(OtherStore))
768 // Otherwise, the other block ended with a conditional branch. If one of the
769 // destinations is StoreBB, then we have the if/then case.
770 if (OtherBr->getSuccessor(0) != StoreBB &&
771 OtherBr->getSuccessor(1) != StoreBB)
774 // Okay, we know that OtherBr now goes to Dest and StoreBB, so this is an
775 // if/then triangle. See if there is a store to the same ptr as SI that
778 // Check to see if we find the matching store.
779 if ((OtherStore = dyn_cast<StoreInst>(BBI))) {
780 if (OtherStore->getOperand(1) != SI.getOperand(1) ||
781 !SI.isSameOperationAs(OtherStore))
785 // If we find something that may be using or overwriting the stored
786 // value, or if we run out of instructions, we can't do the xform.
787 if (BBI->mayReadFromMemory() || BBI->mayWriteToMemory() ||
788 BBI == OtherBB->begin())
792 // In order to eliminate the store in OtherBr, we have to
793 // make sure nothing reads or overwrites the stored value in
795 for (BasicBlock::iterator I = StoreBB->begin(); &*I != &SI; ++I) {
796 // FIXME: This should really be AA driven.
797 if (I->mayReadFromMemory() || I->mayWriteToMemory())
802 // Insert a PHI node now if we need it.
803 Value *MergedVal = OtherStore->getOperand(0);
804 if (MergedVal != SI.getOperand(0)) {
805 PHINode *PN = PHINode::Create(MergedVal->getType(), 2, "storemerge");
806 PN->addIncoming(SI.getOperand(0), SI.getParent());
807 PN->addIncoming(OtherStore->getOperand(0), OtherBB);
808 MergedVal = InsertNewInstBefore(PN, DestBB->front());
811 // Advance to a place where it is safe to insert the new store and
813 BBI = DestBB->getFirstInsertionPt();
814 StoreInst *NewSI = new StoreInst(MergedVal, SI.getOperand(1),
819 InsertNewInstBefore(NewSI, *BBI);
820 NewSI->setDebugLoc(OtherStore->getDebugLoc());
822 // Nuke the old stores.
823 EraseInstFromFunction(SI);
824 EraseInstFromFunction(*OtherStore);