#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/Analysis/AliasAnalysis.h"
+#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/CFG.h"
#include "llvm/Analysis/CaptureTracking.h"
#include "llvm/Analysis/InstructionSimplify.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/MemoryBuiltins.h"
+#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/Operator.h"
#include "llvm/Pass.h"
#include "llvm/Support/ErrorHandling.h"
-#include "llvm/Target/TargetLibraryInfo.h"
#include <algorithm>
using namespace llvm;
+/// Enable analysis of recursive PHI nodes.
+static cl::opt<bool> EnableRecPhiAnalysis("basicaa-recphi",
+ cl::Hidden, cl::init(false));
+
/// Cutoff after which to stop analysing a set of phi nodes potentially involved
/// in a cycle. Because we are analysing 'through' phi nodes we need to be
/// careful with value equivalence. We use reachability to make sure a value
const TargetLibraryInfo &TLI,
bool RoundToAlign = false) {
uint64_t Size;
- if (getObjectSize(V, Size, &DL, &TLI, RoundToAlign))
+ if (getObjectSize(V, Size, DL, &TLI, RoundToAlign))
return Size;
- return AliasAnalysis::UnknownSize;
+ return MemoryLocation::UnknownSize;
}
/// isObjectSmallerThan - Return true if we can prove that the object specified
// reads a bit past the end given sufficient alignment.
uint64_t ObjectSize = getObjectSize(V, DL, TLI, /*RoundToAlign*/true);
- return ObjectSize != AliasAnalysis::UnknownSize && ObjectSize < Size;
+ return ObjectSize != MemoryLocation::UnknownSize && ObjectSize < Size;
}
/// isObjectSize - Return true if we can prove that the object specified
static bool isObjectSize(const Value *V, uint64_t Size,
const DataLayout &DL, const TargetLibraryInfo &TLI) {
uint64_t ObjectSize = getObjectSize(V, DL, TLI);
- return ObjectSize != AliasAnalysis::UnknownSize && ObjectSize == Size;
-}
-
-/// isIdentifiedFunctionLocal - Return true if V is umabigously identified
-/// at the function-level. Different IdentifiedFunctionLocals can't alias.
-/// Further, an IdentifiedFunctionLocal can not alias with any function
-/// arguments other than itself, which is not necessarily true for
-/// IdentifiedObjects.
-static bool isIdentifiedFunctionLocal(const Value *V)
-{
- return isa<AllocaInst>(V) || isNoAliasCall(V) || isNoAliasArgument(V);
+ return ObjectSize != MemoryLocation::UnknownSize && ObjectSize == Size;
}
-
//===----------------------------------------------------------------------===//
// GetElementPtr Instruction Decomposition and Analysis
//===----------------------------------------------------------------------===//
/// represented in the result.
static Value *GetLinearExpression(Value *V, APInt &Scale, APInt &Offset,
ExtensionKind &Extension,
- const DataLayout &DL, unsigned Depth) {
+ const DataLayout &DL, unsigned Depth,
+ AssumptionCache *AC, DominatorTree *DT) {
assert(V->getType()->isIntegerTy() && "Not an integer value");
// Limit our recursion depth.
return V;
}
+ if (ConstantInt *Const = dyn_cast<ConstantInt>(V)) {
+ // if it's a constant, just convert it to an offset
+ // and remove the variable.
+ Offset += Const->getValue();
+ assert(Scale == 0 && "Constant values don't have a scale");
+ return V;
+ }
+
if (BinaryOperator *BOp = dyn_cast<BinaryOperator>(V)) {
if (ConstantInt *RHSC = dyn_cast<ConstantInt>(BOp->getOperand(1))) {
switch (BOp->getOpcode()) {
case Instruction::Or:
// X|C == X+C if all the bits in C are unset in X. Otherwise we can't
// analyze it.
- if (!MaskedValueIsZero(BOp->getOperand(0), RHSC->getValue(), &DL))
+ if (!MaskedValueIsZero(BOp->getOperand(0), RHSC->getValue(), DL, 0, AC,
+ BOp, DT))
break;
// FALL THROUGH.
case Instruction::Add:
V = GetLinearExpression(BOp->getOperand(0), Scale, Offset, Extension,
- DL, Depth+1);
+ DL, Depth + 1, AC, DT);
Offset += RHSC->getValue();
return V;
case Instruction::Mul:
V = GetLinearExpression(BOp->getOperand(0), Scale, Offset, Extension,
- DL, Depth+1);
+ DL, Depth + 1, AC, DT);
Offset *= RHSC->getValue();
Scale *= RHSC->getValue();
return V;
case Instruction::Shl:
V = GetLinearExpression(BOp->getOperand(0), Scale, Offset, Extension,
- DL, Depth+1);
+ DL, Depth + 1, AC, DT);
Offset <<= RHSC->getValue().getLimitedValue();
Scale <<= RHSC->getValue().getLimitedValue();
return V;
Offset = Offset.trunc(SmallWidth);
Extension = isa<SExtInst>(V) ? EK_SignExt : EK_ZeroExt;
- Value *Result = GetLinearExpression(CastOp, Scale, Offset, Extension,
- DL, Depth+1);
+ Value *Result = GetLinearExpression(CastOp, Scale, Offset, Extension, DL,
+ Depth + 1, AC, DT);
Scale = Scale.zext(OldWidth);
- Offset = Offset.zext(OldWidth);
+
+ // We have to sign-extend even if Extension == EK_ZeroExt as we can't
+ // decompose a sign extension (i.e. zext(x - 1) != zext(x) - zext(-1)).
+ Offset = Offset.sext(OldWidth);
return Result;
}
static const Value *
DecomposeGEPExpression(const Value *V, int64_t &BaseOffs,
SmallVectorImpl<VariableGEPIndex> &VarIndices,
- bool &MaxLookupReached, const DataLayout *DL) {
+ bool &MaxLookupReached, const DataLayout &DL,
+ AssumptionCache *AC, DominatorTree *DT) {
// Limit recursion depth to limit compile time in crazy cases.
unsigned MaxLookup = MaxLookupSearchDepth;
MaxLookupReached = false;
return V;
}
- if (Op->getOpcode() == Instruction::BitCast) {
+ if (Op->getOpcode() == Instruction::BitCast ||
+ Op->getOpcode() == Instruction::AddrSpaceCast) {
V = Op->getOperand(0);
continue;
}
// If it's not a GEP, hand it off to SimplifyInstruction to see if it
// can come up with something. This matches what GetUnderlyingObject does.
if (const Instruction *I = dyn_cast<Instruction>(V))
- // TODO: Get a DominatorTree and use it here.
+ // TODO: Get a DominatorTree and AssumptionCache and use them here
+ // (these are both now available in this function, but this should be
+ // updated when GetUnderlyingObject is updated). TLI should be
+ // provided also.
if (const Value *Simplified =
SimplifyInstruction(const_cast<Instruction *>(I), DL)) {
V = Simplified;
if (!GEPOp->getOperand(0)->getType()->getPointerElementType()->isSized())
return V;
- // If we are lacking DataLayout information, we can't compute the offets of
- // elements computed by GEPs. However, we can handle bitcast equivalent
- // GEPs.
- if (!DL) {
- if (!GEPOp->hasAllZeroIndices())
- return V;
- V = GEPOp->getOperand(0);
- continue;
- }
-
unsigned AS = GEPOp->getPointerAddressSpace();
// Walk the indices of the GEP, accumulating them into BaseOff/VarIndices.
gep_type_iterator GTI = gep_type_begin(GEPOp);
unsigned FieldNo = cast<ConstantInt>(Index)->getZExtValue();
if (FieldNo == 0) continue;
- BaseOffs += DL->getStructLayout(STy)->getElementOffset(FieldNo);
+ BaseOffs += DL.getStructLayout(STy)->getElementOffset(FieldNo);
continue;
}
// For an array/pointer, add the element offset, explicitly scaled.
if (ConstantInt *CIdx = dyn_cast<ConstantInt>(Index)) {
if (CIdx->isZero()) continue;
- BaseOffs += DL->getTypeAllocSize(*GTI)*CIdx->getSExtValue();
+ BaseOffs += DL.getTypeAllocSize(*GTI) * CIdx->getSExtValue();
continue;
}
- uint64_t Scale = DL->getTypeAllocSize(*GTI);
+ uint64_t Scale = DL.getTypeAllocSize(*GTI);
ExtensionKind Extension = EK_NotExtended;
// If the integer type is smaller than the pointer size, it is implicitly
// sign extended to pointer size.
unsigned Width = Index->getType()->getIntegerBitWidth();
- if (DL->getPointerSizeInBits(AS) > Width)
+ if (DL.getPointerSizeInBits(AS) > Width)
Extension = EK_SignExt;
// Use GetLinearExpression to decompose the index into a C1*V+C2 form.
APInt IndexScale(Width, 0), IndexOffset(Width, 0);
- Index = GetLinearExpression(Index, IndexScale, IndexOffset, Extension,
- *DL, 0);
+ Index = GetLinearExpression(Index, IndexScale, IndexOffset, Extension, DL,
+ 0, AC, DT);
// The GEP index scale ("Scale") scales C1*V+C2, yielding (C1*V+C2)*Scale.
// This gives us an aggregate computation of (C1*Scale)*V + C2*Scale.
// Make sure that we have a scale that makes sense for this target's
// pointer size.
- if (unsigned ShiftBits = 64 - DL->getPointerSizeInBits(AS)) {
+ if (unsigned ShiftBits = 64 - DL.getPointerSizeInBits(AS)) {
Scale <<= ShiftBits;
Scale = (int64_t)Scale >> ShiftBits;
}
initializeBasicAliasAnalysisPass(*PassRegistry::getPassRegistry());
}
- void initializePass() override {
- InitializeAliasAnalysis(this);
- }
+ bool doInitialization(Module &M) override;
void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.addRequired<AliasAnalysis>();
- AU.addRequired<TargetLibraryInfo>();
+ AU.addRequired<AssumptionCacheTracker>();
+ AU.addRequired<TargetLibraryInfoWrapperPass>();
}
- AliasResult alias(const Location &LocA, const Location &LocB) override {
+ AliasResult alias(const MemoryLocation &LocA,
+ const MemoryLocation &LocB) override {
assert(AliasCache.empty() && "AliasCache must be cleared after use!");
assert(notDifferentParent(LocA.Ptr, LocB.Ptr) &&
"BasicAliasAnalysis doesn't support interprocedural queries.");
- AliasResult Alias = aliasCheck(LocA.Ptr, LocA.Size, LocA.TBAATag,
- LocB.Ptr, LocB.Size, LocB.TBAATag);
+ AliasResult Alias = aliasCheck(LocA.Ptr, LocA.Size, LocA.AATags,
+ LocB.Ptr, LocB.Size, LocB.AATags);
// AliasCache rarely has more than 1 or 2 elements, always use
// shrink_and_clear so it quickly returns to the inline capacity of the
// SmallDenseMap if it ever grows larger.
return Alias;
}
- ModRefResult getModRefInfo(ImmutableCallSite CS,
- const Location &Loc) override;
+ ModRefInfo getModRefInfo(ImmutableCallSite CS,
+ const MemoryLocation &Loc) override;
- ModRefResult getModRefInfo(ImmutableCallSite CS1,
- ImmutableCallSite CS2) override {
- // The AliasAnalysis base class has some smarts, lets use them.
- return AliasAnalysis::getModRefInfo(CS1, CS2);
- }
+ ModRefInfo getModRefInfo(ImmutableCallSite CS1,
+ ImmutableCallSite CS2) override;
/// pointsToConstantMemory - Chase pointers until we find a (constant
/// global) or not.
- bool pointsToConstantMemory(const Location &Loc, bool OrLocal) override;
+ bool pointsToConstantMemory(const MemoryLocation &Loc,
+ bool OrLocal) override;
+
+ /// Get the location associated with a pointer argument of a callsite.
+ ModRefInfo getArgModRefInfo(ImmutableCallSite CS, unsigned ArgIdx) override;
/// getModRefBehavior - Return the behavior when calling the given
/// call site.
- ModRefBehavior getModRefBehavior(ImmutableCallSite CS) override;
+ FunctionModRefBehavior getModRefBehavior(ImmutableCallSite CS) override;
/// getModRefBehavior - Return the behavior when calling the given function.
/// For use when the call site is not known.
- ModRefBehavior getModRefBehavior(const Function *F) override;
+ FunctionModRefBehavior getModRefBehavior(const Function *F) override;
/// getAdjustedAnalysisPointer - This method is used when a pass implements
/// an analysis interface through multiple inheritance. If needed, it
private:
// AliasCache - Track alias queries to guard against recursion.
- typedef std::pair<Location, Location> LocPair;
+ typedef std::pair<MemoryLocation, MemoryLocation> LocPair;
typedef SmallDenseMap<LocPair, AliasResult, 8> AliasCacheTy;
AliasCacheTy AliasCache;
// aliasGEP - Provide a bunch of ad-hoc rules to disambiguate a GEP
// instruction against another.
AliasResult aliasGEP(const GEPOperator *V1, uint64_t V1Size,
- const MDNode *V1TBAAInfo,
+ const AAMDNodes &V1AAInfo,
const Value *V2, uint64_t V2Size,
- const MDNode *V2TBAAInfo,
+ const AAMDNodes &V2AAInfo,
const Value *UnderlyingV1, const Value *UnderlyingV2);
// aliasPHI - Provide a bunch of ad-hoc rules to disambiguate a PHI
// instruction against another.
AliasResult aliasPHI(const PHINode *PN, uint64_t PNSize,
- const MDNode *PNTBAAInfo,
+ const AAMDNodes &PNAAInfo,
const Value *V2, uint64_t V2Size,
- const MDNode *V2TBAAInfo);
+ const AAMDNodes &V2AAInfo);
/// aliasSelect - Disambiguate a Select instruction against another value.
AliasResult aliasSelect(const SelectInst *SI, uint64_t SISize,
- const MDNode *SITBAAInfo,
+ const AAMDNodes &SIAAInfo,
const Value *V2, uint64_t V2Size,
- const MDNode *V2TBAAInfo);
+ const AAMDNodes &V2AAInfo);
AliasResult aliasCheck(const Value *V1, uint64_t V1Size,
- const MDNode *V1TBAATag,
+ AAMDNodes V1AATag,
const Value *V2, uint64_t V2Size,
- const MDNode *V2TBAATag);
+ AAMDNodes V2AATag);
};
} // End of anonymous namespace
INITIALIZE_AG_PASS_BEGIN(BasicAliasAnalysis, AliasAnalysis, "basicaa",
"Basic Alias Analysis (stateless AA impl)",
false, true, false)
-INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfo)
+INITIALIZE_PASS_DEPENDENCY(AssumptionCacheTracker)
+INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfoWrapperPass)
INITIALIZE_AG_PASS_END(BasicAliasAnalysis, AliasAnalysis, "basicaa",
"Basic Alias Analysis (stateless AA impl)",
false, true, false)
/// pointsToConstantMemory - Returns whether the given pointer value
/// points to memory that is local to the function, with global constants being
/// considered local to all functions.
-bool
-BasicAliasAnalysis::pointsToConstantMemory(const Location &Loc, bool OrLocal) {
+bool BasicAliasAnalysis::pointsToConstantMemory(const MemoryLocation &Loc,
+ bool OrLocal) {
assert(Visited.empty() && "Visited must be cleared after use!");
unsigned MaxLookup = 8;
SmallVector<const Value *, 16> Worklist;
Worklist.push_back(Loc.Ptr);
do {
- const Value *V = GetUnderlyingObject(Worklist.pop_back_val(), DL);
- if (!Visited.insert(V)) {
+ const Value *V = GetUnderlyingObject(Worklist.pop_back_val(), *DL);
+ if (!Visited.insert(V).second) {
Visited.clear();
return AliasAnalysis::pointsToConstantMemory(Loc, OrLocal);
}
Visited.clear();
return AliasAnalysis::pointsToConstantMemory(Loc, OrLocal);
}
- for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
- Worklist.push_back(PN->getIncomingValue(i));
+ for (Value *IncValue : PN->incoming_values())
+ Worklist.push_back(IncValue);
continue;
}
return Worklist.empty();
}
+// FIXME: This code is duplicated with MemoryLocation and should be hoisted to
+// some common utility location.
+static bool isMemsetPattern16(const Function *MS,
+ const TargetLibraryInfo &TLI) {
+ if (TLI.has(LibFunc::memset_pattern16) &&
+ MS->getName() == "memset_pattern16") {
+ FunctionType *MemsetType = MS->getFunctionType();
+ if (!MemsetType->isVarArg() && MemsetType->getNumParams() == 3 &&
+ isa<PointerType>(MemsetType->getParamType(0)) &&
+ isa<PointerType>(MemsetType->getParamType(1)) &&
+ isa<IntegerType>(MemsetType->getParamType(2)))
+ return true;
+ }
+
+ return false;
+}
+
/// getModRefBehavior - Return the behavior when calling the given call site.
-AliasAnalysis::ModRefBehavior
+FunctionModRefBehavior
BasicAliasAnalysis::getModRefBehavior(ImmutableCallSite CS) {
if (CS.doesNotAccessMemory())
// Can't do better than this.
- return DoesNotAccessMemory;
+ return FMRB_DoesNotAccessMemory;
- ModRefBehavior Min = UnknownModRefBehavior;
+ FunctionModRefBehavior Min = FMRB_UnknownModRefBehavior;
// If the callsite knows it only reads memory, don't return worse
// than that.
if (CS.onlyReadsMemory())
- Min = OnlyReadsMemory;
+ Min = FMRB_OnlyReadsMemory;
+
+ if (CS.onlyAccessesArgMemory())
+ Min = FunctionModRefBehavior(Min & FMRB_OnlyAccessesArgumentPointees);
// The AliasAnalysis base class has some smarts, lets use them.
- return ModRefBehavior(AliasAnalysis::getModRefBehavior(CS) & Min);
+ return FunctionModRefBehavior(AliasAnalysis::getModRefBehavior(CS) & Min);
}
/// getModRefBehavior - Return the behavior when calling the given function.
/// For use when the call site is not known.
-AliasAnalysis::ModRefBehavior
+FunctionModRefBehavior
BasicAliasAnalysis::getModRefBehavior(const Function *F) {
// If the function declares it doesn't access memory, we can't do better.
if (F->doesNotAccessMemory())
- return DoesNotAccessMemory;
+ return FMRB_DoesNotAccessMemory;
// For intrinsics, we can check the table.
- if (unsigned iid = F->getIntrinsicID()) {
+ if (Intrinsic::ID iid = F->getIntrinsicID()) {
#define GET_INTRINSIC_MODREF_BEHAVIOR
#include "llvm/IR/Intrinsics.gen"
#undef GET_INTRINSIC_MODREF_BEHAVIOR
}
- ModRefBehavior Min = UnknownModRefBehavior;
+ FunctionModRefBehavior Min = FMRB_UnknownModRefBehavior;
// If the function declares it only reads memory, go with that.
if (F->onlyReadsMemory())
- Min = OnlyReadsMemory;
+ Min = FMRB_OnlyReadsMemory;
+
+ if (F->onlyAccessesArgMemory())
+ Min = FunctionModRefBehavior(Min & FMRB_OnlyAccessesArgumentPointees);
+
+ const TargetLibraryInfo &TLI =
+ getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
+ if (isMemsetPattern16(F, TLI))
+ Min = FMRB_OnlyAccessesArgumentPointees;
// Otherwise be conservative.
- return ModRefBehavior(AliasAnalysis::getModRefBehavior(F) & Min);
+ return FunctionModRefBehavior(AliasAnalysis::getModRefBehavior(F) & Min);
+}
+
+ModRefInfo BasicAliasAnalysis::getArgModRefInfo(ImmutableCallSite CS,
+ unsigned ArgIdx) {
+ if (const IntrinsicInst *II = dyn_cast<IntrinsicInst>(CS.getInstruction()))
+ switch (II->getIntrinsicID()) {
+ default:
+ break;
+ case Intrinsic::memset:
+ case Intrinsic::memcpy:
+ case Intrinsic::memmove:
+ assert((ArgIdx == 0 || ArgIdx == 1) &&
+ "Invalid argument index for memory intrinsic");
+ return ArgIdx ? MRI_Ref : MRI_Mod;
+ }
+
+ // We can bound the aliasing properties of memset_pattern16 just as we can
+ // for memcpy/memset. This is particularly important because the
+ // LoopIdiomRecognizer likes to turn loops into calls to memset_pattern16
+ // whenever possible.
+ if (CS.getCalledFunction() &&
+ isMemsetPattern16(CS.getCalledFunction(), *TLI)) {
+ assert((ArgIdx == 0 || ArgIdx == 1) &&
+ "Invalid argument index for memset_pattern16");
+ return ArgIdx ? MRI_Ref : MRI_Mod;
+ }
+ // FIXME: Handle memset_pattern4 and memset_pattern8 also.
+
+ return AliasAnalysis::getArgModRefInfo(CS, ArgIdx);
+}
+
+static bool isAssumeIntrinsic(ImmutableCallSite CS) {
+ const IntrinsicInst *II = dyn_cast<IntrinsicInst>(CS.getInstruction());
+ if (II && II->getIntrinsicID() == Intrinsic::assume)
+ return true;
+
+ return false;
+}
+
+bool BasicAliasAnalysis::doInitialization(Module &M) {
+ InitializeAliasAnalysis(this, &M.getDataLayout());
+ return true;
}
/// getModRefInfo - Check to see if the specified callsite can clobber the
/// specified memory object. Since we only look at local properties of this
/// function, we really can't say much about this query. We do, however, use
/// simple "address taken" analysis on local objects.
-AliasAnalysis::ModRefResult
-BasicAliasAnalysis::getModRefInfo(ImmutableCallSite CS,
- const Location &Loc) {
+ModRefInfo BasicAliasAnalysis::getModRefInfo(ImmutableCallSite CS,
+ const MemoryLocation &Loc) {
assert(notDifferentParent(CS.getInstruction(), Loc.Ptr) &&
"AliasAnalysis query involving multiple functions!");
- const Value *Object = GetUnderlyingObject(Loc.Ptr, DL);
+ const Value *Object = GetUnderlyingObject(Loc.Ptr, *DL);
// If this is a tail call and Loc.Ptr points to a stack location, we know that
// the tail call cannot access or modify the local stack.
if (isa<AllocaInst>(Object))
if (const CallInst *CI = dyn_cast<CallInst>(CS.getInstruction()))
if (CI->isTailCall())
- return NoModRef;
+ return MRI_NoModRef;
// If the pointer is to a locally allocated object that does not escape,
// then the call can not mod/ref the pointer unless the call takes the pointer
// is impossible to alias the pointer we're checking. If not, we have to
// assume that the call could touch the pointer, even though it doesn't
// escape.
- if (!isNoAlias(Location(*CI), Location(Object))) {
+ if (!isNoAlias(MemoryLocation(*CI), MemoryLocation(Object))) {
PassedAsArg = true;
break;
}
}
if (!PassedAsArg)
- return NoModRef;
+ return MRI_NoModRef;
}
- const TargetLibraryInfo &TLI = getAnalysis<TargetLibraryInfo>();
- ModRefResult Min = ModRef;
+ // While the assume intrinsic is marked as arbitrarily writing so that
+ // proper control dependencies will be maintained, it never aliases any
+ // particular memory location.
+ if (isAssumeIntrinsic(CS))
+ return MRI_NoModRef;
- // Finally, handle specific knowledge of intrinsics.
- const IntrinsicInst *II = dyn_cast<IntrinsicInst>(CS.getInstruction());
- if (II != nullptr)
- switch (II->getIntrinsicID()) {
- default: break;
- case Intrinsic::memcpy:
- case Intrinsic::memmove: {
- uint64_t Len = UnknownSize;
- if (ConstantInt *LenCI = dyn_cast<ConstantInt>(II->getArgOperand(2)))
- Len = LenCI->getZExtValue();
- Value *Dest = II->getArgOperand(0);
- Value *Src = II->getArgOperand(1);
- // If it can't overlap the source dest, then it doesn't modref the loc.
- if (isNoAlias(Location(Dest, Len), Loc)) {
- if (isNoAlias(Location(Src, Len), Loc))
- return NoModRef;
- // If it can't overlap the dest, then worst case it reads the loc.
- Min = Ref;
- } else if (isNoAlias(Location(Src, Len), Loc)) {
- // If it can't overlap the source, then worst case it mutates the loc.
- Min = Mod;
- }
- break;
- }
- case Intrinsic::memset:
- // Since memset is 'accesses arguments' only, the AliasAnalysis base class
- // will handle it for the variable length case.
- if (ConstantInt *LenCI = dyn_cast<ConstantInt>(II->getArgOperand(2))) {
- uint64_t Len = LenCI->getZExtValue();
- Value *Dest = II->getArgOperand(0);
- if (isNoAlias(Location(Dest, Len), Loc))
- return NoModRef;
- }
- // We know that memset doesn't load anything.
- Min = Mod;
- break;
- case Intrinsic::lifetime_start:
- case Intrinsic::lifetime_end:
- case Intrinsic::invariant_start: {
- uint64_t PtrSize =
- cast<ConstantInt>(II->getArgOperand(0))->getZExtValue();
- if (isNoAlias(Location(II->getArgOperand(1),
- PtrSize,
- II->getMetadata(LLVMContext::MD_tbaa)),
- Loc))
- return NoModRef;
- break;
- }
- case Intrinsic::invariant_end: {
- uint64_t PtrSize =
- cast<ConstantInt>(II->getArgOperand(1))->getZExtValue();
- if (isNoAlias(Location(II->getArgOperand(2),
- PtrSize,
- II->getMetadata(LLVMContext::MD_tbaa)),
- Loc))
- return NoModRef;
- break;
- }
- case Intrinsic::arm_neon_vld1: {
- // LLVM's vld1 and vst1 intrinsics currently only support a single
- // vector register.
- uint64_t Size =
- DL ? DL->getTypeStoreSize(II->getType()) : UnknownSize;
- if (isNoAlias(Location(II->getArgOperand(0), Size,
- II->getMetadata(LLVMContext::MD_tbaa)),
- Loc))
- return NoModRef;
- break;
- }
- case Intrinsic::arm_neon_vst1: {
- uint64_t Size =
- DL ? DL->getTypeStoreSize(II->getArgOperand(1)->getType()) : UnknownSize;
- if (isNoAlias(Location(II->getArgOperand(0), Size,
- II->getMetadata(LLVMContext::MD_tbaa)),
- Loc))
- return NoModRef;
- break;
- }
- }
+ // The AliasAnalysis base class has some smarts, lets use them.
+ return AliasAnalysis::getModRefInfo(CS, Loc);
+}
- // We can bound the aliasing properties of memset_pattern16 just as we can
- // for memcpy/memset. This is particularly important because the
- // LoopIdiomRecognizer likes to turn loops into calls to memset_pattern16
- // whenever possible.
- else if (TLI.has(LibFunc::memset_pattern16) &&
- CS.getCalledFunction() &&
- CS.getCalledFunction()->getName() == "memset_pattern16") {
- const Function *MS = CS.getCalledFunction();
- FunctionType *MemsetType = MS->getFunctionType();
- if (!MemsetType->isVarArg() && MemsetType->getNumParams() == 3 &&
- isa<PointerType>(MemsetType->getParamType(0)) &&
- isa<PointerType>(MemsetType->getParamType(1)) &&
- isa<IntegerType>(MemsetType->getParamType(2))) {
- uint64_t Len = UnknownSize;
- if (const ConstantInt *LenCI = dyn_cast<ConstantInt>(CS.getArgument(2)))
- Len = LenCI->getZExtValue();
- const Value *Dest = CS.getArgument(0);
- const Value *Src = CS.getArgument(1);
- // If it can't overlap the source dest, then it doesn't modref the loc.
- if (isNoAlias(Location(Dest, Len), Loc)) {
- // Always reads 16 bytes of the source.
- if (isNoAlias(Location(Src, 16), Loc))
- return NoModRef;
- // If it can't overlap the dest, then worst case it reads the loc.
- Min = Ref;
- // Always reads 16 bytes of the source.
- } else if (isNoAlias(Location(Src, 16), Loc)) {
- // If it can't overlap the source, then worst case it mutates the loc.
- Min = Mod;
- }
- }
- }
+ModRefInfo BasicAliasAnalysis::getModRefInfo(ImmutableCallSite CS1,
+ ImmutableCallSite CS2) {
+ // While the assume intrinsic is marked as arbitrarily writing so that
+ // proper control dependencies will be maintained, it never aliases any
+ // particular memory location.
+ if (isAssumeIntrinsic(CS1) || isAssumeIntrinsic(CS2))
+ return MRI_NoModRef;
// The AliasAnalysis base class has some smarts, lets use them.
- return ModRefResult(AliasAnalysis::getModRefInfo(CS, Loc) & Min);
+ return AliasAnalysis::getModRefInfo(CS1, CS2);
+}
+
+/// \brief Provide ad-hoc rules to disambiguate accesses through two GEP
+/// operators, both having the exact same pointer operand.
+static AliasResult aliasSameBasePointerGEPs(const GEPOperator *GEP1,
+ uint64_t V1Size,
+ const GEPOperator *GEP2,
+ uint64_t V2Size,
+ const DataLayout &DL) {
+
+ assert(GEP1->getPointerOperand() == GEP2->getPointerOperand() &&
+ "Expected GEPs with the same pointer operand");
+
+ // Try to determine whether GEP1 and GEP2 index through arrays, into structs,
+ // such that the struct field accesses provably cannot alias.
+ // We also need at least two indices (the pointer, and the struct field).
+ if (GEP1->getNumIndices() != GEP2->getNumIndices() ||
+ GEP1->getNumIndices() < 2)
+ return MayAlias;
+
+ // If we don't know the size of the accesses through both GEPs, we can't
+ // determine whether the struct fields accessed can't alias.
+ if (V1Size == MemoryLocation::UnknownSize ||
+ V2Size == MemoryLocation::UnknownSize)
+ return MayAlias;
+
+ ConstantInt *C1 =
+ dyn_cast<ConstantInt>(GEP1->getOperand(GEP1->getNumOperands() - 1));
+ ConstantInt *C2 =
+ dyn_cast<ConstantInt>(GEP2->getOperand(GEP2->getNumOperands() - 1));
+
+ // If the last (struct) indices aren't constants, we can't say anything.
+ // If they're identical, the other indices might be also be dynamically
+ // equal, so the GEPs can alias.
+ if (!C1 || !C2 || C1 == C2)
+ return MayAlias;
+
+ // Find the last-indexed type of the GEP, i.e., the type you'd get if
+ // you stripped the last index.
+ // On the way, look at each indexed type. If there's something other
+ // than an array, different indices can lead to different final types.
+ SmallVector<Value *, 8> IntermediateIndices;
+
+ // Insert the first index; we don't need to check the type indexed
+ // through it as it only drops the pointer indirection.
+ assert(GEP1->getNumIndices() > 1 && "Not enough GEP indices to examine");
+ IntermediateIndices.push_back(GEP1->getOperand(1));
+
+ // Insert all the remaining indices but the last one.
+ // Also, check that they all index through arrays.
+ for (unsigned i = 1, e = GEP1->getNumIndices() - 1; i != e; ++i) {
+ if (!isa<ArrayType>(GetElementPtrInst::getIndexedType(
+ GEP1->getSourceElementType(), IntermediateIndices)))
+ return MayAlias;
+ IntermediateIndices.push_back(GEP1->getOperand(i + 1));
+ }
+
+ StructType *LastIndexedStruct =
+ dyn_cast<StructType>(GetElementPtrInst::getIndexedType(
+ GEP1->getSourceElementType(), IntermediateIndices));
+
+ if (!LastIndexedStruct)
+ return MayAlias;
+
+ // We know that:
+ // - both GEPs begin indexing from the exact same pointer;
+ // - the last indices in both GEPs are constants, indexing into a struct;
+ // - said indices are different, hence, the pointed-to fields are different;
+ // - both GEPs only index through arrays prior to that.
+ //
+ // This lets us determine that the struct that GEP1 indexes into and the
+ // struct that GEP2 indexes into must either precisely overlap or be
+ // completely disjoint. Because they cannot partially overlap, indexing into
+ // different non-overlapping fields of the struct will never alias.
+
+ // Therefore, the only remaining thing needed to show that both GEPs can't
+ // alias is that the fields are not overlapping.
+ const StructLayout *SL = DL.getStructLayout(LastIndexedStruct);
+ const uint64_t StructSize = SL->getSizeInBytes();
+ const uint64_t V1Off = SL->getElementOffset(C1->getZExtValue());
+ const uint64_t V2Off = SL->getElementOffset(C2->getZExtValue());
+
+ auto EltsDontOverlap = [StructSize](uint64_t V1Off, uint64_t V1Size,
+ uint64_t V2Off, uint64_t V2Size) {
+ return V1Off < V2Off && V1Off + V1Size <= V2Off &&
+ ((V2Off + V2Size <= StructSize) ||
+ (V2Off + V2Size - StructSize <= V1Off));
+ };
+
+ if (EltsDontOverlap(V1Off, V1Size, V2Off, V2Size) ||
+ EltsDontOverlap(V2Off, V2Size, V1Off, V1Size))
+ return NoAlias;
+
+ return MayAlias;
}
/// aliasGEP - Provide a bunch of ad-hoc rules to disambiguate a GEP instruction
/// anything about V2. UnderlyingV1 is GetUnderlyingObject(GEP1, DL),
/// UnderlyingV2 is the same for V2.
///
-AliasAnalysis::AliasResult
-BasicAliasAnalysis::aliasGEP(const GEPOperator *GEP1, uint64_t V1Size,
- const MDNode *V1TBAAInfo,
- const Value *V2, uint64_t V2Size,
- const MDNode *V2TBAAInfo,
- const Value *UnderlyingV1,
- const Value *UnderlyingV2) {
+AliasResult BasicAliasAnalysis::aliasGEP(
+ const GEPOperator *GEP1, uint64_t V1Size, const AAMDNodes &V1AAInfo,
+ const Value *V2, uint64_t V2Size, const AAMDNodes &V2AAInfo,
+ const Value *UnderlyingV1, const Value *UnderlyingV2) {
int64_t GEP1BaseOffset;
bool GEP1MaxLookupReached;
SmallVector<VariableGEPIndex, 4> GEP1VariableIndices;
+ // We have to get two AssumptionCaches here because GEP1 and V2 may be from
+ // different functions.
+ // FIXME: This really doesn't make any sense. We get a dominator tree below
+ // that can only refer to a single function. But this function (aliasGEP) is
+ // a method on an immutable pass that can be called when there *isn't*
+ // a single function. The old pass management layer makes this "work", but
+ // this isn't really a clean solution.
+ AssumptionCacheTracker &ACT = getAnalysis<AssumptionCacheTracker>();
+ AssumptionCache *AC1 = nullptr, *AC2 = nullptr;
+ if (auto *GEP1I = dyn_cast<Instruction>(GEP1))
+ AC1 = &ACT.getAssumptionCache(
+ const_cast<Function &>(*GEP1I->getParent()->getParent()));
+ if (auto *I2 = dyn_cast<Instruction>(V2))
+ AC2 = &ACT.getAssumptionCache(
+ const_cast<Function &>(*I2->getParent()->getParent()));
+
+ DominatorTreeWrapperPass *DTWP =
+ getAnalysisIfAvailable<DominatorTreeWrapperPass>();
+ DominatorTree *DT = DTWP ? &DTWP->getDomTree() : nullptr;
+
// If we have two gep instructions with must-alias or not-alias'ing base
// pointers, figure out if the indexes to the GEP tell us anything about the
// derived pointer.
if (const GEPOperator *GEP2 = dyn_cast<GEPOperator>(V2)) {
// Do the base pointers alias?
- AliasResult BaseAlias = aliasCheck(UnderlyingV1, UnknownSize, nullptr,
- UnderlyingV2, UnknownSize, nullptr);
+ AliasResult BaseAlias =
+ aliasCheck(UnderlyingV1, MemoryLocation::UnknownSize, AAMDNodes(),
+ UnderlyingV2, MemoryLocation::UnknownSize, AAMDNodes());
// Check for geps of non-aliasing underlying pointers where the offsets are
// identical.
if ((BaseAlias == MayAlias) && V1Size == V2Size) {
// Do the base pointers alias assuming type and size.
AliasResult PreciseBaseAlias = aliasCheck(UnderlyingV1, V1Size,
- V1TBAAInfo, UnderlyingV2,
- V2Size, V2TBAAInfo);
+ V1AAInfo, UnderlyingV2,
+ V2Size, V2AAInfo);
if (PreciseBaseAlias == NoAlias) {
// See if the computed offset from the common pointer tells us about the
// relation of the resulting pointer.
bool GEP2MaxLookupReached;
SmallVector<VariableGEPIndex, 4> GEP2VariableIndices;
const Value *GEP2BasePtr =
- DecomposeGEPExpression(GEP2, GEP2BaseOffset, GEP2VariableIndices,
- GEP2MaxLookupReached, DL);
+ DecomposeGEPExpression(GEP2, GEP2BaseOffset, GEP2VariableIndices,
+ GEP2MaxLookupReached, *DL, AC2, DT);
const Value *GEP1BasePtr =
- DecomposeGEPExpression(GEP1, GEP1BaseOffset, GEP1VariableIndices,
- GEP1MaxLookupReached, DL);
+ DecomposeGEPExpression(GEP1, GEP1BaseOffset, GEP1VariableIndices,
+ GEP1MaxLookupReached, *DL, AC1, DT);
// DecomposeGEPExpression and GetUnderlyingObject should return the
// same result except when DecomposeGEPExpression has no DataLayout.
if (GEP1BasePtr != UnderlyingV1 || GEP2BasePtr != UnderlyingV2) {
// exactly, see if the computed offset from the common pointer tells us
// about the relation of the resulting pointer.
const Value *GEP1BasePtr =
- DecomposeGEPExpression(GEP1, GEP1BaseOffset, GEP1VariableIndices,
- GEP1MaxLookupReached, DL);
+ DecomposeGEPExpression(GEP1, GEP1BaseOffset, GEP1VariableIndices,
+ GEP1MaxLookupReached, *DL, AC1, DT);
int64_t GEP2BaseOffset;
bool GEP2MaxLookupReached;
SmallVector<VariableGEPIndex, 4> GEP2VariableIndices;
const Value *GEP2BasePtr =
- DecomposeGEPExpression(GEP2, GEP2BaseOffset, GEP2VariableIndices,
- GEP2MaxLookupReached, DL);
+ DecomposeGEPExpression(GEP2, GEP2BaseOffset, GEP2VariableIndices,
+ GEP2MaxLookupReached, *DL, AC2, DT);
// DecomposeGEPExpression and GetUnderlyingObject should return the
// same result except when DecomposeGEPExpression has no DataLayout.
"DecomposeGEPExpression and GetUnderlyingObject disagree!");
return MayAlias;
}
+
+ // If we know the two GEPs are based off of the exact same pointer (and not
+ // just the same underlying object), see if that tells us anything about
+ // the resulting pointers.
+ if (DL && GEP1->getPointerOperand() == GEP2->getPointerOperand()) {
+ AliasResult R = aliasSameBasePointerGEPs(GEP1, V1Size, GEP2, V2Size, *DL);
+ // If we couldn't find anything interesting, don't abandon just yet.
+ if (R != MayAlias)
+ return R;
+ }
+
// If the max search depth is reached the result is undefined
if (GEP2MaxLookupReached || GEP1MaxLookupReached)
return MayAlias;
// pointer, we know they cannot alias.
// If both accesses are unknown size, we can't do anything useful here.
- if (V1Size == UnknownSize && V2Size == UnknownSize)
+ if (V1Size == MemoryLocation::UnknownSize &&
+ V2Size == MemoryLocation::UnknownSize)
return MayAlias;
- AliasResult R = aliasCheck(UnderlyingV1, UnknownSize, nullptr,
- V2, V2Size, V2TBAAInfo);
+ AliasResult R = aliasCheck(UnderlyingV1, MemoryLocation::UnknownSize,
+ AAMDNodes(), V2, V2Size, V2AAInfo);
if (R != MustAlias)
// If V2 may alias GEP base pointer, conservatively returns MayAlias.
// If V2 is known not to alias GEP base pointer, then the two values
return R;
const Value *GEP1BasePtr =
- DecomposeGEPExpression(GEP1, GEP1BaseOffset, GEP1VariableIndices,
- GEP1MaxLookupReached, DL);
+ DecomposeGEPExpression(GEP1, GEP1BaseOffset, GEP1VariableIndices,
+ GEP1MaxLookupReached, *DL, AC1, DT);
// DecomposeGEPExpression and GetUnderlyingObject should return the
// same result except when DecomposeGEPExpression has no DataLayout.
// greater, we know they do not overlap.
if (GEP1BaseOffset != 0 && GEP1VariableIndices.empty()) {
if (GEP1BaseOffset >= 0) {
- if (V2Size != UnknownSize) {
+ if (V2Size != MemoryLocation::UnknownSize) {
if ((uint64_t)GEP1BaseOffset < V2Size)
return PartialAlias;
return NoAlias;
// GEP1 V2
// We need to know that V2Size is not unknown, otherwise we might have
// stripped a gep with negative index ('gep <ptr>, -1, ...).
- if (V1Size != UnknownSize && V2Size != UnknownSize) {
+ if (V1Size != MemoryLocation::UnknownSize &&
+ V2Size != MemoryLocation::UnknownSize) {
if (-(uint64_t)GEP1BaseOffset < V1Size)
return PartialAlias;
return NoAlias;
}
}
- // Try to distinguish something like &A[i][1] against &A[42][0].
- // Grab the least significant bit set in any of the scales.
if (!GEP1VariableIndices.empty()) {
uint64_t Modulo = 0;
- for (unsigned i = 0, e = GEP1VariableIndices.size(); i != e; ++i)
- Modulo |= (uint64_t)GEP1VariableIndices[i].Scale;
+ bool AllPositive = true;
+ for (unsigned i = 0, e = GEP1VariableIndices.size(); i != e; ++i) {
+
+ // Try to distinguish something like &A[i][1] against &A[42][0].
+ // Grab the least significant bit set in any of the scales. We
+ // don't need std::abs here (even if the scale's negative) as we'll
+ // be ^'ing Modulo with itself later.
+ Modulo |= (uint64_t) GEP1VariableIndices[i].Scale;
+
+ if (AllPositive) {
+ // If the Value could change between cycles, then any reasoning about
+ // the Value this cycle may not hold in the next cycle. We'll just
+ // give up if we can't determine conditions that hold for every cycle:
+ const Value *V = GEP1VariableIndices[i].V;
+
+ bool SignKnownZero, SignKnownOne;
+ ComputeSignBit(const_cast<Value *>(V), SignKnownZero, SignKnownOne, *DL,
+ 0, AC1, nullptr, DT);
+
+ // Zero-extension widens the variable, and so forces the sign
+ // bit to zero.
+ bool IsZExt = GEP1VariableIndices[i].Extension == EK_ZeroExt;
+ SignKnownZero |= IsZExt;
+ SignKnownOne &= !IsZExt;
+
+ // If the variable begins with a zero then we know it's
+ // positive, regardless of whether the value is signed or
+ // unsigned.
+ int64_t Scale = GEP1VariableIndices[i].Scale;
+ AllPositive =
+ (SignKnownZero && Scale >= 0) ||
+ (SignKnownOne && Scale < 0);
+ }
+ }
+
Modulo = Modulo ^ (Modulo & (Modulo - 1));
// We can compute the difference between the two addresses
// mod Modulo. Check whether that difference guarantees that the
// two locations do not alias.
uint64_t ModOffset = (uint64_t)GEP1BaseOffset & (Modulo - 1);
- if (V1Size != UnknownSize && V2Size != UnknownSize &&
- ModOffset >= V2Size && V1Size <= Modulo - ModOffset)
+ if (V1Size != MemoryLocation::UnknownSize &&
+ V2Size != MemoryLocation::UnknownSize && ModOffset >= V2Size &&
+ V1Size <= Modulo - ModOffset)
+ return NoAlias;
+
+ // If we know all the variables are positive, then GEP1 >= GEP1BasePtr.
+ // If GEP1BasePtr > V2 (GEP1BaseOffset > 0) then we know the pointers
+ // don't alias if V2Size can fit in the gap between V2 and GEP1BasePtr.
+ if (AllPositive && GEP1BaseOffset > 0 && V2Size <= (uint64_t) GEP1BaseOffset)
return NoAlias;
}
return PartialAlias;
}
-static AliasAnalysis::AliasResult
-MergeAliasResults(AliasAnalysis::AliasResult A, AliasAnalysis::AliasResult B) {
+static AliasResult MergeAliasResults(AliasResult A, AliasResult B) {
// If the results agree, take it.
if (A == B)
return A;
// A mix of PartialAlias and MustAlias is PartialAlias.
- if ((A == AliasAnalysis::PartialAlias && B == AliasAnalysis::MustAlias) ||
- (B == AliasAnalysis::PartialAlias && A == AliasAnalysis::MustAlias))
- return AliasAnalysis::PartialAlias;
+ if ((A == PartialAlias && B == MustAlias) ||
+ (B == PartialAlias && A == MustAlias))
+ return PartialAlias;
// Otherwise, we don't know anything.
- return AliasAnalysis::MayAlias;
+ return MayAlias;
}
/// aliasSelect - Provide a bunch of ad-hoc rules to disambiguate a Select
/// instruction against another.
-AliasAnalysis::AliasResult
-BasicAliasAnalysis::aliasSelect(const SelectInst *SI, uint64_t SISize,
- const MDNode *SITBAAInfo,
- const Value *V2, uint64_t V2Size,
- const MDNode *V2TBAAInfo) {
+AliasResult BasicAliasAnalysis::aliasSelect(const SelectInst *SI,
+ uint64_t SISize,
+ const AAMDNodes &SIAAInfo,
+ const Value *V2, uint64_t V2Size,
+ const AAMDNodes &V2AAInfo) {
// If the values are Selects with the same condition, we can do a more precise
// check: just check for aliases between the values on corresponding arms.
if (const SelectInst *SI2 = dyn_cast<SelectInst>(V2))
if (SI->getCondition() == SI2->getCondition()) {
AliasResult Alias =
- aliasCheck(SI->getTrueValue(), SISize, SITBAAInfo,
- SI2->getTrueValue(), V2Size, V2TBAAInfo);
+ aliasCheck(SI->getTrueValue(), SISize, SIAAInfo,
+ SI2->getTrueValue(), V2Size, V2AAInfo);
if (Alias == MayAlias)
return MayAlias;
AliasResult ThisAlias =
- aliasCheck(SI->getFalseValue(), SISize, SITBAAInfo,
- SI2->getFalseValue(), V2Size, V2TBAAInfo);
+ aliasCheck(SI->getFalseValue(), SISize, SIAAInfo,
+ SI2->getFalseValue(), V2Size, V2AAInfo);
return MergeAliasResults(ThisAlias, Alias);
}
// If both arms of the Select node NoAlias or MustAlias V2, then returns
// NoAlias / MustAlias. Otherwise, returns MayAlias.
AliasResult Alias =
- aliasCheck(V2, V2Size, V2TBAAInfo, SI->getTrueValue(), SISize, SITBAAInfo);
+ aliasCheck(V2, V2Size, V2AAInfo, SI->getTrueValue(), SISize, SIAAInfo);
if (Alias == MayAlias)
return MayAlias;
AliasResult ThisAlias =
- aliasCheck(V2, V2Size, V2TBAAInfo, SI->getFalseValue(), SISize, SITBAAInfo);
+ aliasCheck(V2, V2Size, V2AAInfo, SI->getFalseValue(), SISize, SIAAInfo);
return MergeAliasResults(ThisAlias, Alias);
}
// aliasPHI - Provide a bunch of ad-hoc rules to disambiguate a PHI instruction
// against another.
-AliasAnalysis::AliasResult
-BasicAliasAnalysis::aliasPHI(const PHINode *PN, uint64_t PNSize,
- const MDNode *PNTBAAInfo,
- const Value *V2, uint64_t V2Size,
- const MDNode *V2TBAAInfo) {
+AliasResult BasicAliasAnalysis::aliasPHI(const PHINode *PN, uint64_t PNSize,
+ const AAMDNodes &PNAAInfo,
+ const Value *V2, uint64_t V2Size,
+ const AAMDNodes &V2AAInfo) {
// Track phi nodes we have visited. We use this information when we determine
// value equivalence.
VisitedPhiBBs.insert(PN->getParent());
// on corresponding edges.
if (const PHINode *PN2 = dyn_cast<PHINode>(V2))
if (PN2->getParent() == PN->getParent()) {
- LocPair Locs(Location(PN, PNSize, PNTBAAInfo),
- Location(V2, V2Size, V2TBAAInfo));
+ LocPair Locs(MemoryLocation(PN, PNSize, PNAAInfo),
+ MemoryLocation(V2, V2Size, V2AAInfo));
if (PN > V2)
std::swap(Locs.first, Locs.second);
// Analyse the PHIs' inputs under the assumption that the PHIs are
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
AliasResult ThisAlias =
- aliasCheck(PN->getIncomingValue(i), PNSize, PNTBAAInfo,
+ aliasCheck(PN->getIncomingValue(i), PNSize, PNAAInfo,
PN2->getIncomingValueForBlock(PN->getIncomingBlock(i)),
- V2Size, V2TBAAInfo);
+ V2Size, V2AAInfo);
Alias = MergeAliasResults(ThisAlias, Alias);
if (Alias == MayAlias)
break;
SmallPtrSet<Value*, 4> UniqueSrc;
SmallVector<Value*, 4> V1Srcs;
- for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
- Value *PV1 = PN->getIncomingValue(i);
+ bool isRecursive = false;
+ for (Value *PV1 : PN->incoming_values()) {
if (isa<PHINode>(PV1))
// If any of the source itself is a PHI, return MayAlias conservatively
// to avoid compile time explosion. The worst possible case is if both
// sides are PHI nodes. In which case, this is O(m x n) time where 'm'
// and 'n' are the number of PHI sources.
return MayAlias;
- if (UniqueSrc.insert(PV1))
+
+ if (EnableRecPhiAnalysis)
+ if (GEPOperator *PV1GEP = dyn_cast<GEPOperator>(PV1)) {
+ // Check whether the incoming value is a GEP that advances the pointer
+ // result of this PHI node (e.g. in a loop). If this is the case, we
+ // would recurse and always get a MayAlias. Handle this case specially
+ // below.
+ if (PV1GEP->getPointerOperand() == PN && PV1GEP->getNumIndices() == 1 &&
+ isa<ConstantInt>(PV1GEP->idx_begin())) {
+ isRecursive = true;
+ continue;
+ }
+ }
+
+ if (UniqueSrc.insert(PV1).second)
V1Srcs.push_back(PV1);
}
- AliasResult Alias = aliasCheck(V2, V2Size, V2TBAAInfo,
- V1Srcs[0], PNSize, PNTBAAInfo);
+ // If this PHI node is recursive, set the size of the accessed memory to
+ // unknown to represent all the possible values the GEP could advance the
+ // pointer to.
+ if (isRecursive)
+ PNSize = MemoryLocation::UnknownSize;
+
+ AliasResult Alias = aliasCheck(V2, V2Size, V2AAInfo,
+ V1Srcs[0], PNSize, PNAAInfo);
+
// Early exit if the check of the first PHI source against V2 is MayAlias.
// Other results are not possible.
if (Alias == MayAlias)
for (unsigned i = 1, e = V1Srcs.size(); i != e; ++i) {
Value *V = V1Srcs[i];
- AliasResult ThisAlias = aliasCheck(V2, V2Size, V2TBAAInfo,
- V, PNSize, PNTBAAInfo);
+ AliasResult ThisAlias = aliasCheck(V2, V2Size, V2AAInfo,
+ V, PNSize, PNAAInfo);
Alias = MergeAliasResults(ThisAlias, Alias);
if (Alias == MayAlias)
break;
// aliasCheck - Provide a bunch of ad-hoc rules to disambiguate in common cases,
// such as array references.
//
-AliasAnalysis::AliasResult
-BasicAliasAnalysis::aliasCheck(const Value *V1, uint64_t V1Size,
- const MDNode *V1TBAAInfo,
- const Value *V2, uint64_t V2Size,
- const MDNode *V2TBAAInfo) {
+AliasResult BasicAliasAnalysis::aliasCheck(const Value *V1, uint64_t V1Size,
+ AAMDNodes V1AAInfo, const Value *V2,
+ uint64_t V2Size,
+ AAMDNodes V2AAInfo) {
// If either of the memory references is empty, it doesn't matter what the
// pointer values are.
if (V1Size == 0 || V2Size == 0)
V1 = V1->stripPointerCasts();
V2 = V2->stripPointerCasts();
+ // If V1 or V2 is undef, the result is NoAlias because we can always pick a
+ // value for undef that aliases nothing in the program.
+ if (isa<UndefValue>(V1) || isa<UndefValue>(V2))
+ return NoAlias;
+
// Are we checking for alias of the same value?
// Because we look 'through' phi nodes we could look at "Value" pointers from
// different iterations. We must therefore make sure that this is not the
return NoAlias; // Scalars cannot alias each other
// Figure out what objects these things are pointing to if we can.
- const Value *O1 = GetUnderlyingObject(V1, DL, MaxLookupSearchDepth);
- const Value *O2 = GetUnderlyingObject(V2, DL, MaxLookupSearchDepth);
+ const Value *O1 = GetUnderlyingObject(V1, *DL, MaxLookupSearchDepth);
+ const Value *O2 = GetUnderlyingObject(V2, *DL, MaxLookupSearchDepth);
// Null values in the default address space don't point to any object, so they
// don't alias any other pointer.
// If the size of one access is larger than the entire object on the other
// side, then we know such behavior is undefined and can assume no alias.
if (DL)
- if ((V1Size != UnknownSize && isObjectSmallerThan(O2, V1Size, *DL, *TLI)) ||
- (V2Size != UnknownSize && isObjectSmallerThan(O1, V2Size, *DL, *TLI)))
+ if ((V1Size != MemoryLocation::UnknownSize &&
+ isObjectSmallerThan(O2, V1Size, *DL, *TLI)) ||
+ (V2Size != MemoryLocation::UnknownSize &&
+ isObjectSmallerThan(O1, V2Size, *DL, *TLI)))
return NoAlias;
// Check the cache before climbing up use-def chains. This also terminates
// otherwise infinitely recursive queries.
- LocPair Locs(Location(V1, V1Size, V1TBAAInfo),
- Location(V2, V2Size, V2TBAAInfo));
+ LocPair Locs(MemoryLocation(V1, V1Size, V1AAInfo),
+ MemoryLocation(V2, V2Size, V2AAInfo));
if (V1 > V2)
std::swap(Locs.first, Locs.second);
std::pair<AliasCacheTy::iterator, bool> Pair =
std::swap(V1, V2);
std::swap(V1Size, V2Size);
std::swap(O1, O2);
- std::swap(V1TBAAInfo, V2TBAAInfo);
+ std::swap(V1AAInfo, V2AAInfo);
}
if (const GEPOperator *GV1 = dyn_cast<GEPOperator>(V1)) {
- AliasResult Result = aliasGEP(GV1, V1Size, V1TBAAInfo, V2, V2Size, V2TBAAInfo, O1, O2);
+ AliasResult Result = aliasGEP(GV1, V1Size, V1AAInfo, V2, V2Size, V2AAInfo, O1, O2);
if (Result != MayAlias) return AliasCache[Locs] = Result;
}
if (isa<PHINode>(V2) && !isa<PHINode>(V1)) {
std::swap(V1, V2);
std::swap(V1Size, V2Size);
- std::swap(V1TBAAInfo, V2TBAAInfo);
+ std::swap(V1AAInfo, V2AAInfo);
}
if (const PHINode *PN = dyn_cast<PHINode>(V1)) {
- AliasResult Result = aliasPHI(PN, V1Size, V1TBAAInfo,
- V2, V2Size, V2TBAAInfo);
+ AliasResult Result = aliasPHI(PN, V1Size, V1AAInfo,
+ V2, V2Size, V2AAInfo);
if (Result != MayAlias) return AliasCache[Locs] = Result;
}
if (isa<SelectInst>(V2) && !isa<SelectInst>(V1)) {
std::swap(V1, V2);
std::swap(V1Size, V2Size);
- std::swap(V1TBAAInfo, V2TBAAInfo);
+ std::swap(V1AAInfo, V2AAInfo);
}
if (const SelectInst *S1 = dyn_cast<SelectInst>(V1)) {
- AliasResult Result = aliasSelect(S1, V1Size, V1TBAAInfo,
- V2, V2Size, V2TBAAInfo);
+ AliasResult Result = aliasSelect(S1, V1Size, V1AAInfo,
+ V2, V2Size, V2AAInfo);
if (Result != MayAlias) return AliasCache[Locs] = Result;
}
// accesses is accessing the entire object, then the accesses must
// overlap in some way.
if (DL && O1 == O2)
- if ((V1Size != UnknownSize && isObjectSize(O1, V1Size, *DL, *TLI)) ||
- (V2Size != UnknownSize && isObjectSize(O2, V2Size, *DL, *TLI)))
+ if ((V1Size != MemoryLocation::UnknownSize &&
+ isObjectSize(O1, V1Size, *DL, *TLI)) ||
+ (V2Size != MemoryLocation::UnknownSize &&
+ isObjectSize(O2, V2Size, *DL, *TLI)))
return AliasCache[Locs] = PartialAlias;
AliasResult Result =
- AliasAnalysis::alias(Location(V1, V1Size, V1TBAAInfo),
- Location(V2, V2Size, V2TBAAInfo));
+ AliasAnalysis::alias(MemoryLocation(V1, V1Size, V1AAInfo),
+ MemoryLocation(V2, V2Size, V2AAInfo));
return AliasCache[Locs] = Result;
}
if (!Inst)
return true;
+ if (VisitedPhiBBs.empty())
+ return true;
+
if (VisitedPhiBBs.size() > MaxNumPhiBBsValueReachabilityCheck)
return false;
DominatorTreeWrapperPass *DTWP =
getAnalysisIfAvailable<DominatorTreeWrapperPass>();
DominatorTree *DT = DTWP ? &DTWP->getDomTree() : nullptr;
- LoopInfo *LI = getAnalysisIfAvailable<LoopInfo>();
+ auto *LIWP = getAnalysisIfAvailable<LoopInfoWrapperPass>();
+ LoopInfo *LI = LIWP ? &LIWP->getLoopInfo() : nullptr;
// Make sure that the visited phis cannot reach the Value. This ensures that
// the Values cannot come from different iterations of a potential cycle the
// phi nodes could be involved in.
- for (SmallPtrSet<const BasicBlock *, 8>::iterator PI = VisitedPhiBBs.begin(),
- PE = VisitedPhiBBs.end();
- PI != PE; ++PI)
- if (isPotentiallyReachable((*PI)->begin(), Inst, DT, LI))
+ for (auto *P : VisitedPhiBBs)
+ if (isPotentiallyReachable(P->begin(), Inst, DT, LI))
return false;
return true;
projects / oota-llvm.git / blobdiff