-namespace {
-/// \brief Analyses memory accesses in a loop.
-///
-/// Checks whether run time pointer checks are needed and builds sets for data
-/// dependence checking.
-class AccessAnalysis {
-public:
- /// \brief Read or write access location.
- typedef PointerIntPair<Value *, 1, bool> MemAccessInfo;
- typedef SmallPtrSet<MemAccessInfo, 8> MemAccessInfoSet;
-
- /// \brief Set of potential dependent memory accesses.
- typedef EquivalenceClasses<MemAccessInfo> DepCandidates;
-
- AccessAnalysis(const DataLayout *Dl, AliasAnalysis *AA, DepCandidates &DA) :
- DL(Dl), AST(*AA), DepCands(DA), IsRTCheckNeeded(false) {}
-
- /// \brief Register a load and whether it is only read from.
- void addLoad(AliasAnalysis::Location &Loc, bool IsReadOnly) {
- Value *Ptr = const_cast<Value*>(Loc.Ptr);
- AST.add(Ptr, AliasAnalysis::UnknownSize, Loc.AATags);
- Accesses.insert(MemAccessInfo(Ptr, false));
- if (IsReadOnly)
- ReadOnlyPtr.insert(Ptr);
- }
-
- /// \brief Register a store.
- void addStore(AliasAnalysis::Location &Loc) {
- Value *Ptr = const_cast<Value*>(Loc.Ptr);
- AST.add(Ptr, AliasAnalysis::UnknownSize, Loc.AATags);
- Accesses.insert(MemAccessInfo(Ptr, true));
- }
-
- /// \brief Check whether we can check the pointers at runtime for
- /// non-intersection.
- bool canCheckPtrAtRT(LoopVectorizationLegality::RuntimePointerCheck &RtCheck,
- unsigned &NumComparisons, ScalarEvolution *SE,
- Loop *TheLoop, ValueToValueMap &Strides,
- bool ShouldCheckStride = false);
-
- /// \brief Goes over all memory accesses, checks whether a RT check is needed
- /// and builds sets of dependent accesses.
- void buildDependenceSets() {
- processMemAccesses();
- }
-
- bool isRTCheckNeeded() { return IsRTCheckNeeded; }
-
- bool isDependencyCheckNeeded() { return !CheckDeps.empty(); }
- void resetDepChecks() { CheckDeps.clear(); }
-
- MemAccessInfoSet &getDependenciesToCheck() { return CheckDeps; }
-
-private:
- typedef SetVector<MemAccessInfo> PtrAccessSet;
-
- /// \brief Go over all memory access and check whether runtime pointer checks
- /// are needed /// and build sets of dependency check candidates.
- void processMemAccesses();
-
- /// Set of all accesses.
- PtrAccessSet Accesses;
-
- /// Set of accesses that need a further dependence check.
- MemAccessInfoSet CheckDeps;
-
- /// Set of pointers that are read only.
- SmallPtrSet<Value*, 16> ReadOnlyPtr;
-
- const DataLayout *DL;
-
- /// An alias set tracker to partition the access set by underlying object and
- //intrinsic property (such as TBAA metadata).
- AliasSetTracker AST;
-
- /// Sets of potentially dependent accesses - members of one set share an
- /// underlying pointer. The set "CheckDeps" identfies which sets really need a
- /// dependence check.
- DepCandidates &DepCands;
-
- bool IsRTCheckNeeded;
-};
-
-} // end anonymous namespace
-
-/// \brief Check whether a pointer can participate in a runtime bounds check.
-static bool hasComputableBounds(ScalarEvolution *SE, ValueToValueMap &Strides,
- Value *Ptr) {
- const SCEV *PtrScev = replaceSymbolicStrideSCEV(SE, Strides, Ptr);
- const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(PtrScev);
- if (!AR)
- return false;
-
- return AR->isAffine();
-}
-
-/// \brief Check the stride of the pointer and ensure that it does not wrap in
-/// the address space.
-static int isStridedPtr(ScalarEvolution *SE, const DataLayout *DL, Value *Ptr,
- const Loop *Lp, ValueToValueMap &StridesMap);
-
-bool AccessAnalysis::canCheckPtrAtRT(
- LoopVectorizationLegality::RuntimePointerCheck &RtCheck,
- unsigned &NumComparisons, ScalarEvolution *SE, Loop *TheLoop,
- ValueToValueMap &StridesMap, bool ShouldCheckStride) {
- // Find pointers with computable bounds. We are going to use this information
- // to place a runtime bound check.
- bool CanDoRT = true;
-
- bool IsDepCheckNeeded = isDependencyCheckNeeded();
- NumComparisons = 0;
-
- // We assign a consecutive id to access from different alias sets.
- // Accesses between different groups doesn't need to be checked.
- unsigned ASId = 1;
- for (auto &AS : AST) {
- unsigned NumReadPtrChecks = 0;
- unsigned NumWritePtrChecks = 0;
-
- // We assign consecutive id to access from different dependence sets.
- // Accesses within the same set don't need a runtime check.
- unsigned RunningDepId = 1;
- DenseMap<Value *, unsigned> DepSetId;
-
- for (auto A : AS) {
- Value *Ptr = A.getValue();
- bool IsWrite = Accesses.count(MemAccessInfo(Ptr, true));
- MemAccessInfo Access(Ptr, IsWrite);
-
- if (IsWrite)
- ++NumWritePtrChecks;
- else
- ++NumReadPtrChecks;
-
- if (hasComputableBounds(SE, StridesMap, Ptr) &&
- // When we run after a failing dependency check we have to make sure we
- // don't have wrapping pointers.
- (!ShouldCheckStride ||
- isStridedPtr(SE, DL, Ptr, TheLoop, StridesMap) == 1)) {
- // The id of the dependence set.
- unsigned DepId;
-
- if (IsDepCheckNeeded) {
- Value *Leader = DepCands.getLeaderValue(Access).getPointer();
- unsigned &LeaderId = DepSetId[Leader];
- if (!LeaderId)
- LeaderId = RunningDepId++;
- DepId = LeaderId;
- } else
- // Each access has its own dependence set.
- DepId = RunningDepId++;
-
- RtCheck.insert(SE, TheLoop, Ptr, IsWrite, DepId, ASId, StridesMap);
-
- DEBUG(dbgs() << "LV: Found a runtime check ptr:" << *Ptr << '\n');
- } else {
- CanDoRT = false;
- }
- }
-
- if (IsDepCheckNeeded && CanDoRT && RunningDepId == 2)
- NumComparisons += 0; // Only one dependence set.
- else {
- NumComparisons += (NumWritePtrChecks * (NumReadPtrChecks +
- NumWritePtrChecks - 1));
- }
-
- ++ASId;
- }
-
- // If the pointers that we would use for the bounds comparison have different
- // address spaces, assume the values aren't directly comparable, so we can't
- // use them for the runtime check. We also have to assume they could
- // overlap. In the future there should be metadata for whether address spaces
- // are disjoint.
- unsigned NumPointers = RtCheck.Pointers.size();
- for (unsigned i = 0; i < NumPointers; ++i) {
- for (unsigned j = i + 1; j < NumPointers; ++j) {
- // Only need to check pointers between two different dependency sets.
- if (RtCheck.DependencySetId[i] == RtCheck.DependencySetId[j])
- continue;
- // Only need to check pointers in the same alias set.
- if (RtCheck.AliasSetId[i] != RtCheck.AliasSetId[j])
- continue;
-
- Value *PtrI = RtCheck.Pointers[i];
- Value *PtrJ = RtCheck.Pointers[j];
-
- unsigned ASi = PtrI->getType()->getPointerAddressSpace();
- unsigned ASj = PtrJ->getType()->getPointerAddressSpace();
- if (ASi != ASj) {
- DEBUG(dbgs() << "LV: Runtime check would require comparison between"
- " different address spaces\n");
- return false;
- }
- }
- }
-
- return CanDoRT;
-}
-
-void AccessAnalysis::processMemAccesses() {
- // We process the set twice: first we process read-write pointers, last we
- // process read-only pointers. This allows us to skip dependence tests for
- // read-only pointers.
-
- DEBUG(dbgs() << "LV: Processing memory accesses...\n");
- DEBUG(dbgs() << " AST: "; AST.dump());
- DEBUG(dbgs() << "LV: Accesses:\n");
- DEBUG({
- for (auto A : Accesses)
- dbgs() << "\t" << *A.getPointer() << " (" <<
- (A.getInt() ? "write" : (ReadOnlyPtr.count(A.getPointer()) ?
- "read-only" : "read")) << ")\n";
- });
-
- // The AliasSetTracker has nicely partitioned our pointers by metadata
- // compatibility and potential for underlying-object overlap. As a result, we
- // only need to check for potential pointer dependencies within each alias
- // set.
- for (auto &AS : AST) {
- // Note that both the alias-set tracker and the alias sets themselves used
- // linked lists internally and so the iteration order here is deterministic
- // (matching the original instruction order within each set).
-
- bool SetHasWrite = false;
-
- // Map of pointers to last access encountered.
- typedef DenseMap<Value*, MemAccessInfo> UnderlyingObjToAccessMap;
- UnderlyingObjToAccessMap ObjToLastAccess;
-
- // Set of access to check after all writes have been processed.
- PtrAccessSet DeferredAccesses;
-
- // Iterate over each alias set twice, once to process read/write pointers,
- // and then to process read-only pointers.
- for (int SetIteration = 0; SetIteration < 2; ++SetIteration) {
- bool UseDeferred = SetIteration > 0;
- PtrAccessSet &S = UseDeferred ? DeferredAccesses : Accesses;
-
- for (auto A : AS) {
- Value *Ptr = A.getValue();
- bool IsWrite = S.count(MemAccessInfo(Ptr, true));
-
- // If we're using the deferred access set, then it contains only reads.
- bool IsReadOnlyPtr = ReadOnlyPtr.count(Ptr) && !IsWrite;
- if (UseDeferred && !IsReadOnlyPtr)
- continue;
- // Otherwise, the pointer must be in the PtrAccessSet, either as a read
- // or a write.
- assert(((IsReadOnlyPtr && UseDeferred) || IsWrite ||
- S.count(MemAccessInfo(Ptr, false))) &&
- "Alias-set pointer not in the access set?");
-
- MemAccessInfo Access(Ptr, IsWrite);
- DepCands.insert(Access);
-
- // Memorize read-only pointers for later processing and skip them in the
- // first round (they need to be checked after we have seen all write
- // pointers). Note: we also mark pointer that are not consecutive as
- // "read-only" pointers (so that we check "a[b[i]] +="). Hence, we need
- // the second check for "!IsWrite".
- if (!UseDeferred && IsReadOnlyPtr) {
- DeferredAccesses.insert(Access);
- continue;
- }
-
- // If this is a write - check other reads and writes for conflicts. If
- // this is a read only check other writes for conflicts (but only if
- // there is no other write to the ptr - this is an optimization to
- // catch "a[i] = a[i] + " without having to do a dependence check).
- if ((IsWrite || IsReadOnlyPtr) && SetHasWrite) {
- CheckDeps.insert(Access);
- IsRTCheckNeeded = true;
- }
-
- if (IsWrite)
- SetHasWrite = true;
-
- // Create sets of pointers connected by a shared alias set and
- // underlying object.
- typedef SmallVector<Value *, 16> ValueVector;
- ValueVector TempObjects;
- GetUnderlyingObjects(Ptr, TempObjects, DL);
- for (Value *UnderlyingObj : TempObjects) {
- UnderlyingObjToAccessMap::iterator Prev =
- ObjToLastAccess.find(UnderlyingObj);
- if (Prev != ObjToLastAccess.end())
- DepCands.unionSets(Access, Prev->second);
-
- ObjToLastAccess[UnderlyingObj] = Access;
- }
- }
- }
- }
-}
-
-namespace {
-/// \brief Checks memory dependences among accesses to the same underlying
-/// object to determine whether there vectorization is legal or not (and at
-/// which vectorization factor).
-///
-/// This class works under the assumption that we already checked that memory
-/// locations with different underlying pointers are "must-not alias".
-/// We use the ScalarEvolution framework to symbolically evalutate access
-/// functions pairs. Since we currently don't restructure the loop we can rely
-/// on the program order of memory accesses to determine their safety.
-/// At the moment we will only deem accesses as safe for:
-/// * A negative constant distance assuming program order.
-///
-/// Safe: tmp = a[i + 1]; OR a[i + 1] = x;
-/// a[i] = tmp; y = a[i];
-///
-/// The latter case is safe because later checks guarantuee that there can't
-/// be a cycle through a phi node (that is, we check that "x" and "y" is not
-/// the same variable: a header phi can only be an induction or a reduction, a
-/// reduction can't have a memory sink, an induction can't have a memory
-/// source). This is important and must not be violated (or we have to
-/// resort to checking for cycles through memory).
-///
-/// * A positive constant distance assuming program order that is bigger
-/// than the biggest memory access.
-///
-/// tmp = a[i] OR b[i] = x
-/// a[i+2] = tmp y = b[i+2];
-///
-/// Safe distance: 2 x sizeof(a[0]), and 2 x sizeof(b[0]), respectively.
-///
-/// * Zero distances and all accesses have the same size.
-///
-class MemoryDepChecker {
-public:
- typedef PointerIntPair<Value *, 1, bool> MemAccessInfo;
- typedef SmallPtrSet<MemAccessInfo, 8> MemAccessInfoSet;
-
- MemoryDepChecker(ScalarEvolution *Se, const DataLayout *Dl, const Loop *L)
- : SE(Se), DL(Dl), InnermostLoop(L), AccessIdx(0),
- ShouldRetryWithRuntimeCheck(false) {}
-
- /// \brief Register the location (instructions are given increasing numbers)
- /// of a write access.
- void addAccess(StoreInst *SI) {
- Value *Ptr = SI->getPointerOperand();
- Accesses[MemAccessInfo(Ptr, true)].push_back(AccessIdx);
- InstMap.push_back(SI);
- ++AccessIdx;
- }
-
- /// \brief Register the location (instructions are given increasing numbers)
- /// of a write access.
- void addAccess(LoadInst *LI) {
- Value *Ptr = LI->getPointerOperand();
- Accesses[MemAccessInfo(Ptr, false)].push_back(AccessIdx);
- InstMap.push_back(LI);
- ++AccessIdx;
- }
-
- /// \brief Check whether the dependencies between the accesses are safe.
- ///
- /// Only checks sets with elements in \p CheckDeps.
- bool areDepsSafe(AccessAnalysis::DepCandidates &AccessSets,
- MemAccessInfoSet &CheckDeps, ValueToValueMap &Strides);
-
- /// \brief The maximum number of bytes of a vector register we can vectorize
- /// the accesses safely with.
- unsigned getMaxSafeDepDistBytes() { return MaxSafeDepDistBytes; }
-
- /// \brief In same cases when the dependency check fails we can still
- /// vectorize the loop with a dynamic array access check.
- bool shouldRetryWithRuntimeCheck() { return ShouldRetryWithRuntimeCheck; }
-
-private:
- ScalarEvolution *SE;
- const DataLayout *DL;
- const Loop *InnermostLoop;
-
- /// \brief Maps access locations (ptr, read/write) to program order.
- DenseMap<MemAccessInfo, std::vector<unsigned> > Accesses;
-
- /// \brief Memory access instructions in program order.
- SmallVector<Instruction *, 16> InstMap;
-
- /// \brief The program order index to be used for the next instruction.
- unsigned AccessIdx;
-
- // We can access this many bytes in parallel safely.
- unsigned MaxSafeDepDistBytes;
-
- /// \brief If we see a non-constant dependence distance we can still try to
- /// vectorize this loop with runtime checks.
- bool ShouldRetryWithRuntimeCheck;
-
- /// \brief Check whether there is a plausible dependence between the two
- /// accesses.
- ///
- /// Access \p A must happen before \p B in program order. The two indices
- /// identify the index into the program order map.
- ///
- /// This function checks whether there is a plausible dependence (or the
- /// absence of such can't be proved) between the two accesses. If there is a
- /// plausible dependence but the dependence distance is bigger than one
- /// element access it records this distance in \p MaxSafeDepDistBytes (if this
- /// distance is smaller than any other distance encountered so far).
- /// Otherwise, this function returns true signaling a possible dependence.
- bool isDependent(const MemAccessInfo &A, unsigned AIdx,
- const MemAccessInfo &B, unsigned BIdx,
- ValueToValueMap &Strides);
-
- /// \brief Check whether the data dependence could prevent store-load
- /// forwarding.
- bool couldPreventStoreLoadForward(unsigned Distance, unsigned TypeByteSize);
-};
-
-} // end anonymous namespace
-
-static bool isInBoundsGep(Value *Ptr) {
- if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Ptr))
- return GEP->isInBounds();
- return false;
-}
-
-/// \brief Check whether the access through \p Ptr has a constant stride.
-static int isStridedPtr(ScalarEvolution *SE, const DataLayout *DL, Value *Ptr,
- const Loop *Lp, ValueToValueMap &StridesMap) {
- const Type *Ty = Ptr->getType();
- assert(Ty->isPointerTy() && "Unexpected non-ptr");
-
- // Make sure that the pointer does not point to aggregate types.
- const PointerType *PtrTy = cast<PointerType>(Ty);
- if (PtrTy->getElementType()->isAggregateType()) {
- DEBUG(dbgs() << "LV: Bad stride - Not a pointer to a scalar type" << *Ptr <<
- "\n");
- return 0;
- }
-
- const SCEV *PtrScev = replaceSymbolicStrideSCEV(SE, StridesMap, Ptr);
-
- const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(PtrScev);
- if (!AR) {
- DEBUG(dbgs() << "LV: Bad stride - Not an AddRecExpr pointer "
- << *Ptr << " SCEV: " << *PtrScev << "\n");
- return 0;
- }
-
- // The accesss function must stride over the innermost loop.
- if (Lp != AR->getLoop()) {
- DEBUG(dbgs() << "LV: Bad stride - Not striding over innermost loop " <<
- *Ptr << " SCEV: " << *PtrScev << "\n");
- }
-
- // The address calculation must not wrap. Otherwise, a dependence could be
- // inverted.
- // An inbounds getelementptr that is a AddRec with a unit stride
- // cannot wrap per definition. The unit stride requirement is checked later.
- // An getelementptr without an inbounds attribute and unit stride would have
- // to access the pointer value "0" which is undefined behavior in address
- // space 0, therefore we can also vectorize this case.
- bool IsInBoundsGEP = isInBoundsGep(Ptr);
- bool IsNoWrapAddRec = AR->getNoWrapFlags(SCEV::NoWrapMask);
- bool IsInAddressSpaceZero = PtrTy->getAddressSpace() == 0;
- if (!IsNoWrapAddRec && !IsInBoundsGEP && !IsInAddressSpaceZero) {
- DEBUG(dbgs() << "LV: Bad stride - Pointer may wrap in the address space "
- << *Ptr << " SCEV: " << *PtrScev << "\n");
- return 0;
- }
-
- // Check the step is constant.
- const SCEV *Step = AR->getStepRecurrence(*SE);
-
- // Calculate the pointer stride and check if it is consecutive.
- const SCEVConstant *C = dyn_cast<SCEVConstant>(Step);
- if (!C) {
- DEBUG(dbgs() << "LV: Bad stride - Not a constant strided " << *Ptr <<
- " SCEV: " << *PtrScev << "\n");
- return 0;
- }
-
- int64_t Size = DL->getTypeAllocSize(PtrTy->getElementType());
- const APInt &APStepVal = C->getValue()->getValue();
-
- // Huge step value - give up.
- if (APStepVal.getBitWidth() > 64)
- return 0;
-
- int64_t StepVal = APStepVal.getSExtValue();
-
- // Strided access.
- int64_t Stride = StepVal / Size;
- int64_t Rem = StepVal % Size;
- if (Rem)
- return 0;
-
- // If the SCEV could wrap but we have an inbounds gep with a unit stride we
- // know we can't "wrap around the address space". In case of address space
- // zero we know that this won't happen without triggering undefined behavior.
- if (!IsNoWrapAddRec && (IsInBoundsGEP || IsInAddressSpaceZero) &&
- Stride != 1 && Stride != -1)
- return 0;
-
- return Stride;
-}
-
-bool MemoryDepChecker::couldPreventStoreLoadForward(unsigned Distance,
- unsigned TypeByteSize) {
- // If loads occur at a distance that is not a multiple of a feasible vector
- // factor store-load forwarding does not take place.
- // Positive dependences might cause troubles because vectorizing them might
- // prevent store-load forwarding making vectorized code run a lot slower.
- // a[i] = a[i-3] ^ a[i-8];
- // The stores to a[i:i+1] don't align with the stores to a[i-3:i-2] and
- // hence on your typical architecture store-load forwarding does not take
- // place. Vectorizing in such cases does not make sense.
- // Store-load forwarding distance.
- const unsigned NumCyclesForStoreLoadThroughMemory = 8*TypeByteSize;
- // Maximum vector factor.
- unsigned MaxVFWithoutSLForwardIssues = MaxVectorWidth*TypeByteSize;
- if(MaxSafeDepDistBytes < MaxVFWithoutSLForwardIssues)
- MaxVFWithoutSLForwardIssues = MaxSafeDepDistBytes;
-
- for (unsigned vf = 2*TypeByteSize; vf <= MaxVFWithoutSLForwardIssues;
- vf *= 2) {
- if (Distance % vf && Distance / vf < NumCyclesForStoreLoadThroughMemory) {
- MaxVFWithoutSLForwardIssues = (vf >>=1);
- break;
- }
- }
-
- if (MaxVFWithoutSLForwardIssues< 2*TypeByteSize) {
- DEBUG(dbgs() << "LV: Distance " << Distance <<
- " that could cause a store-load forwarding conflict\n");
- return true;
- }
-
- if (MaxVFWithoutSLForwardIssues < MaxSafeDepDistBytes &&
- MaxVFWithoutSLForwardIssues != MaxVectorWidth*TypeByteSize)
- MaxSafeDepDistBytes = MaxVFWithoutSLForwardIssues;
- return false;
-}
-
-bool MemoryDepChecker::isDependent(const MemAccessInfo &A, unsigned AIdx,
- const MemAccessInfo &B, unsigned BIdx,
- ValueToValueMap &Strides) {
- assert (AIdx < BIdx && "Must pass arguments in program order");
-
- Value *APtr = A.getPointer();
- Value *BPtr = B.getPointer();
- bool AIsWrite = A.getInt();
- bool BIsWrite = B.getInt();
-
- // Two reads are independent.
- if (!AIsWrite && !BIsWrite)