// Variable uniformity checks are inspired by:
// Karrenberg, R. and Hack, S. Whole Function Vectorization.
//
+// The interleaved access vectorization is based on the paper:
+// Dorit Nuzman, Ira Rosen and Ayal Zaks. Auto-Vectorization of Interleaved
+// Data for SIMD
+//
// Other ideas/concepts are from:
// A. Zaks and D. Nuzman. Autovectorization in GCC-two years later.
//
#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/BlockFrequencyInfo.h"
#include "llvm/Analysis/CodeMetrics.h"
+#include "llvm/Analysis/LoopAccessAnalysis.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/LoopIterator.h"
#include "llvm/Analysis/LoopPass.h"
#include "llvm/Transforms/Scalar.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/Local.h"
-#include "llvm/Transforms/Utils/VectorUtils.h"
+#include "llvm/Analysis/VectorUtils.h"
+#include "llvm/Transforms/Utils/LoopUtils.h"
#include <algorithm>
#include <map>
#include <tuple>
STATISTIC(LoopsVectorized, "Number of loops vectorized");
STATISTIC(LoopsAnalyzed, "Number of loops analyzed for vectorization");
-static cl::opt<unsigned>
-VectorizationFactor("force-vector-width", cl::init(0), cl::Hidden,
- cl::desc("Sets the SIMD width. Zero is autoselect."));
-
-static cl::opt<unsigned>
-VectorizationInterleave("force-vector-interleave", cl::init(0), cl::Hidden,
- cl::desc("Sets the vectorization interleave count. "
- "Zero is autoselect."));
-
static cl::opt<bool>
EnableIfConversion("enable-if-conversion", cl::init(true), cl::Hidden,
cl::desc("Enable if-conversion during vectorization."));
"enable-mem-access-versioning", cl::init(true), cl::Hidden,
cl::desc("Enable symblic stride memory access versioning"));
-/// We don't unroll loops with a known constant trip count below this number.
-static const unsigned TinyTripCountUnrollThreshold = 128;
+static cl::opt<bool> EnableInterleavedMemAccesses(
+ "enable-interleaved-mem-accesses", cl::init(false), cl::Hidden,
+ cl::desc("Enable vectorization on interleaved memory accesses in a loop"));
-/// When performing memory disambiguation checks at runtime do not make more
-/// than this number of comparisons.
-static const unsigned RuntimeMemoryCheckThreshold = 8;
+/// Maximum factor for an interleaved memory access.
+static cl::opt<unsigned> MaxInterleaveGroupFactor(
+ "max-interleave-group-factor", cl::Hidden,
+ cl::desc("Maximum factor for an interleaved access group (default = 8)"),
+ cl::init(8));
-/// Maximum simd width.
-static const unsigned MaxVectorWidth = 64;
+/// We don't interleave loops with a known constant trip count below this
+/// number.
+static const unsigned TinyTripCountInterleaveThreshold = 128;
static cl::opt<unsigned> ForceTargetNumScalarRegs(
"force-target-num-scalar-regs", cl::init(0), cl::Hidden,
static cl::opt<unsigned> SmallLoopCost(
"small-loop-cost", cl::init(20), cl::Hidden,
- cl::desc("The cost of a loop that is considered 'small' by the unroller."));
+ cl::desc(
+ "The cost of a loop that is considered 'small' by the interleaver."));
static cl::opt<bool> LoopVectorizeWithBlockFrequency(
"loop-vectorize-with-block-frequency", cl::init(false), cl::Hidden,
"heuristics minimizing code growth in cold regions and being more "
"aggressive in hot regions."));
-// Runtime unroll loops for load/store throughput.
-static cl::opt<bool> EnableLoadStoreRuntimeUnroll(
- "enable-loadstore-runtime-unroll", cl::init(true), cl::Hidden,
- cl::desc("Enable runtime unrolling until load/store ports are saturated"));
+// Runtime interleave loops for load/store throughput.
+static cl::opt<bool> EnableLoadStoreRuntimeInterleave(
+ "enable-loadstore-runtime-interleave", cl::init(true), cl::Hidden,
+ cl::desc(
+ "Enable runtime interleaving until load/store ports are saturated"));
/// The number of stores in a loop that are allowed to need predication.
static cl::opt<unsigned> NumberOfStoresToPredicate(
static cl::opt<bool> EnableIndVarRegisterHeur(
"enable-ind-var-reg-heur", cl::init(true), cl::Hidden,
- cl::desc("Count the induction variable only once when unrolling"));
+ cl::desc("Count the induction variable only once when interleaving"));
static cl::opt<bool> EnableCondStoresVectorization(
"enable-cond-stores-vec", cl::init(false), cl::Hidden,
cl::desc("Enable if predication of stores during vectorization."));
-static cl::opt<unsigned> MaxNestedScalarReductionUF(
- "max-nested-scalar-reduction-unroll", cl::init(2), cl::Hidden,
- cl::desc("The maximum unroll factor to use when unrolling a scalar "
+static cl::opt<unsigned> MaxNestedScalarReductionIC(
+ "max-nested-scalar-reduction-interleave", cl::init(2), cl::Hidden,
+ cl::desc("The maximum interleave count to use when interleaving a scalar "
"reduction in a nested loop."));
namespace {
// Forward declarations.
+class LoopVectorizeHints;
class LoopVectorizationLegality;
class LoopVectorizationCostModel;
-class LoopVectorizeHints;
-
-/// Optimization analysis message produced during vectorization. Messages inform
-/// the user why vectorization did not occur.
-class VectorizationReport {
- std::string Message;
- raw_string_ostream Out;
- Instruction *Instr;
+class LoopVectorizationRequirements;
+/// \brief This modifies LoopAccessReport to initialize message with
+/// loop-vectorizer-specific part.
+class VectorizationReport : public LoopAccessReport {
public:
- VectorizationReport(Instruction *I = nullptr) : Out(Message), Instr(I) {
- Out << "loop not vectorized: ";
- }
-
- template <typename A> VectorizationReport &operator<<(const A &Value) {
- Out << Value;
- return *this;
- }
-
- Instruction *getInstr() { return Instr; }
-
- std::string &str() { return Out.str(); }
- operator Twine() { return Out.str(); }
-
- /// \brief Emit an analysis note with the debug location from the instruction
- /// in \p Message if available. Otherwise use the location of \p TheLoop.
- static void emitAnalysis(VectorizationReport &Message,
- const Function *TheFunction,
- const Loop *TheLoop);
+ VectorizationReport(Instruction *I = nullptr)
+ : LoopAccessReport("loop not vectorized: ", I) {}
+
+ /// \brief This allows promotion of the loop-access analysis report into the
+ /// loop-vectorizer report. It modifies the message to add the
+ /// loop-vectorizer-specific part of the message.
+ explicit VectorizationReport(const LoopAccessReport &R)
+ : LoopAccessReport(Twine("loop not vectorized: ") + R.str(),
+ R.getInstr()) {}
};
+/// A helper function for converting Scalar types to vector types.
+/// If the incoming type is void, we return void. If the VF is 1, we return
+/// the scalar type.
+static Type* ToVectorTy(Type *Scalar, unsigned VF) {
+ if (Scalar->isVoidTy() || VF == 1)
+ return Scalar;
+ return VectorType::get(Scalar, VF);
+}
+
/// InnerLoopVectorizer vectorizes loops which contain only one basic
/// block to a specified vectorization factor (VF).
/// This class performs the widening of scalars into vectors, or multiple
class InnerLoopVectorizer {
public:
InnerLoopVectorizer(Loop *OrigLoop, ScalarEvolution *SE, LoopInfo *LI,
- DominatorTree *DT, const DataLayout *DL,
- const TargetLibraryInfo *TLI, unsigned VecWidth,
+ DominatorTree *DT, const TargetLibraryInfo *TLI,
+ const TargetTransformInfo *TTI, unsigned VecWidth,
unsigned UnrollFactor)
- : OrigLoop(OrigLoop), SE(SE), LI(LI), DT(DT), DL(DL), TLI(TLI),
+ : OrigLoop(OrigLoop), SE(SE), LI(LI), DT(DT), TLI(TLI), TTI(TTI),
VF(VecWidth), UF(UnrollFactor), Builder(SE->getContext()),
Induction(nullptr), OldInduction(nullptr), WidenMap(UnrollFactor),
- Legal(nullptr) {}
+ Legal(nullptr), AddedSafetyChecks(false) {}
// Perform the actual loop widening (vectorization).
void vectorize(LoopVectorizationLegality *L) {
updateAnalysis();
}
+ // Return true if any runtime check is added.
+ bool IsSafetyChecksAdded() {
+ return AddedSafetyChecks;
+ }
+
virtual ~InnerLoopVectorizer() {}
protected:
/// originated from one scalar instruction.
typedef SmallVector<Value*, 2> VectorParts;
- // When we if-convert we need create edge masks. We have to cache values so
- // that we don't end up with exponential recursion/IR.
+ // When we if-convert we need to create edge masks. We have to cache values
+ // so that we don't end up with exponential recursion/IR.
typedef DenseMap<std::pair<BasicBlock*, BasicBlock*>,
VectorParts> EdgeMaskCache;
- /// \brief Add code that checks at runtime if the accessed arrays overlap.
- ///
- /// Returns a pair of instructions where the first element is the first
- /// instruction generated in possibly a sequence of instructions and the
- /// second value is the final comparator value or NULL if no check is needed.
- std::pair<Instruction *, Instruction *> addRuntimeCheck(Instruction *Loc);
-
- /// \brief Add checks for strides that where assumed to be 1.
+ /// \brief Add checks for strides that were assumed to be 1.
///
/// Returns the last check instruction and the first check instruction in the
/// pair as (first, last).
/// broadcast them into a vector.
VectorParts &getVectorValue(Value *V);
+ /// Try to vectorize the interleaved access group that \p Instr belongs to.
+ void vectorizeInterleaveGroup(Instruction *Instr);
+
/// Generate a shuffle sequence that will reverse the vector Vec.
virtual Value *reverseVector(Value *Vec);
DominatorTree *DT;
/// Alias Analysis.
AliasAnalysis *AA;
- /// Data Layout.
- const DataLayout *DL;
/// Target Library Info.
const TargetLibraryInfo *TLI;
+ /// Target Transform Info.
+ const TargetTransformInfo *TTI;
/// The vectorization SIMD factor to use. Each vector will have this many
/// vector elements.
EdgeMaskCache MaskCache;
LoopVectorizationLegality *Legal;
+
+ // Record whether runtime check is added.
+ bool AddedSafetyChecks;
};
class InnerLoopUnroller : public InnerLoopVectorizer {
public:
InnerLoopUnroller(Loop *OrigLoop, ScalarEvolution *SE, LoopInfo *LI,
- DominatorTree *DT, const DataLayout *DL,
- const TargetLibraryInfo *TLI, unsigned UnrollFactor) :
- InnerLoopVectorizer(OrigLoop, SE, LI, DT, DL, TLI, 1, UnrollFactor) { }
+ DominatorTree *DT, const TargetLibraryInfo *TLI,
+ const TargetTransformInfo *TTI, unsigned UnrollFactor)
+ : InnerLoopVectorizer(OrigLoop, SE, LI, DT, TLI, TTI, 1, UnrollFactor) {}
private:
void scalarizeInstruction(Instruction *Instr,
std::string Result;
if (L) {
raw_string_ostream OS(Result);
- const DebugLoc LoopDbgLoc = L->getStartLoc();
- if (!LoopDbgLoc.isUnknown())
- LoopDbgLoc.print(L->getHeader()->getContext(), OS);
+ if (const DebugLoc LoopDbgLoc = L->getStartLoc())
+ LoopDbgLoc.print(OS);
else
// Just print the module name.
OS << L->getHeader()->getParent()->getParent()->getModuleIdentifier();
}
}
-void VectorizationReport::emitAnalysis(VectorizationReport &Message,
- const Function *TheFunction,
- const Loop *TheLoop) {
- DebugLoc DL = TheLoop->getStartLoc();
- if (Instruction *I = Message.getInstr())
- DL = I->getDebugLoc();
- emitOptimizationRemarkAnalysis(TheFunction->getContext(), DEBUG_TYPE,
- *TheFunction, DL, Message.str());
-}
-
/// \brief Propagate known metadata from one instruction to a vector of others.
static void propagateMetadata(SmallVectorImpl<Value *> &To, const Instruction *From) {
for (Value *V : To)
propagateMetadata(I, From);
}
-namespace {
-/// This struct holds information about the memory runtime legality
-/// check that a group of pointers do not overlap.
-struct RuntimePointerCheck {
- RuntimePointerCheck() : Need(false) {}
-
- /// Reset the state of the pointer runtime information.
- void reset() {
- Need = false;
- Pointers.clear();
- Starts.clear();
- Ends.clear();
- IsWritePtr.clear();
- DependencySetId.clear();
- AliasSetId.clear();
- }
-
- /// Insert a pointer and calculate the start and end SCEVs.
- void insert(ScalarEvolution *SE, Loop *Lp, Value *Ptr, bool WritePtr,
- unsigned DepSetId, unsigned ASId, ValueToValueMap &Strides);
-
- /// This flag indicates if we need to add the runtime check.
- bool Need;
- /// Holds the pointers that we need to check.
- SmallVector<TrackingVH<Value>, 2> Pointers;
- /// Holds the pointer value at the beginning of the loop.
- SmallVector<const SCEV*, 2> Starts;
- /// Holds the pointer value at the end of the loop.
- SmallVector<const SCEV*, 2> Ends;
- /// Holds the information if this pointer is used for writing to memory.
- SmallVector<bool, 2> IsWritePtr;
- /// Holds the id of the set of pointers that could be dependent because of a
- /// shared underlying object.
- SmallVector<unsigned, 2> DependencySetId;
- /// Holds the id of the disjoint alias set to which this pointer belongs.
- SmallVector<unsigned, 2> AliasSetId;
-};
-
-/// \brief Drive the analysis of memory accesses in the loop
+/// \brief The group of interleaved loads/stores sharing the same stride and
+/// close to each other.
+///
+/// Each member in this group has an index starting from 0, and the largest
+/// index should be less than interleaved factor, which is equal to the absolute
+/// value of the access's stride.
///
-/// This class is responsible for analyzing the memory accesses of a loop. It
-/// collects the accesses and then its main helper the AccessAnalysis class
-/// finds and categorizes the dependences in buildDependenceSets.
+/// E.g. An interleaved load group of factor 4:
+/// for (unsigned i = 0; i < 1024; i+=4) {
+/// a = A[i]; // Member of index 0
+/// b = A[i+1]; // Member of index 1
+/// d = A[i+3]; // Member of index 3
+/// ...
+/// }
///
-/// For memory dependences that can be analyzed at compile time, it determines
-/// whether the dependence is part of cycle inhibiting vectorization. This work
-/// is delegated to the MemoryDepChecker class.
+/// An interleaved store group of factor 4:
+/// for (unsigned i = 0; i < 1024; i+=4) {
+/// ...
+/// A[i] = a; // Member of index 0
+/// A[i+1] = b; // Member of index 1
+/// A[i+2] = c; // Member of index 2
+/// A[i+3] = d; // Member of index 3
+/// }
///
-/// For memory dependences that cannot be determined at compile time, it
-/// generates run-time checks to prove independence. This is done by
-/// AccessAnalysis::canCheckPtrAtRT and the checks are maintained by the
-/// RuntimePointerCheck class.
-class LoopAccessAnalysis {
+/// Note: the interleaved load group could have gaps (missing members), but
+/// the interleaved store group doesn't allow gaps.
+class InterleaveGroup {
public:
- /// \brief Collection of parameters used from the vectorizer.
- struct VectorizerParams {
- /// \brief Maximum simd width.
- unsigned MaxVectorWidth;
-
- /// \brief VF as overridden by the user.
- unsigned VectorizationFactor;
- /// \brief Interleave factor as overridden by the user.
- unsigned VectorizationInterleave;
-
- /// \\brief When performing memory disambiguation checks at runtime do not
- /// make more than this number of comparisons.
- unsigned RuntimeMemoryCheckThreshold;
-
- VectorizerParams(unsigned MaxVectorWidth,
- unsigned VectorizationFactor,
- unsigned VectorizationInterleave,
- unsigned RuntimeMemoryCheckThreshold) :
- MaxVectorWidth(MaxVectorWidth),
- VectorizationFactor(VectorizationFactor),
- VectorizationInterleave(VectorizationInterleave),
- RuntimeMemoryCheckThreshold(RuntimeMemoryCheckThreshold) {}
- };
+ InterleaveGroup(Instruction *Instr, int Stride, unsigned Align)
+ : Align(Align), SmallestKey(0), LargestKey(0), InsertPos(Instr) {
+ assert(Align && "The alignment should be non-zero");
- LoopAccessAnalysis(Function *F, Loop *L, ScalarEvolution *SE,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
- AliasAnalysis *AA, DominatorTree *DT,
- const VectorizerParams &VectParams) :
- TheFunction(F), TheLoop(L), SE(SE), DL(DL), TLI(TLI), AA(AA), DT(DT),
- NumLoads(0), NumStores(0), MaxSafeDepDistBytes(-1U),
- VectParams(VectParams) {}
+ Factor = std::abs(Stride);
+ assert(Factor > 1 && "Invalid interleave factor");
- /// Return true we can analyze the memory accesses in the loop and there are
- /// no memory dependence cycles. Replaces symbolic strides using Strides.
- bool canVectorizeMemory(ValueToValueMap &Strides);
+ Reverse = Stride < 0;
+ Members[0] = Instr;
+ }
- RuntimePointerCheck *getRuntimePointerCheck() { return &PtrRtCheck; }
+ bool isReverse() const { return Reverse; }
+ unsigned getFactor() const { return Factor; }
+ unsigned getAlignment() const { return Align; }
+ unsigned getNumMembers() const { return Members.size(); }
- /// Return true if the block BB needs to be predicated in order for the loop
- /// to be vectorized.
- bool blockNeedsPredication(BasicBlock *BB);
+ /// \brief Try to insert a new member \p Instr with index \p Index and
+ /// alignment \p NewAlign. The index is related to the leader and it could be
+ /// negative if it is the new leader.
+ ///
+ /// \returns false if the instruction doesn't belong to the group.
+ bool insertMember(Instruction *Instr, int Index, unsigned NewAlign) {
+ assert(NewAlign && "The new member's alignment should be non-zero");
- /// Returns true if the value V is uniform within the loop.
- bool isUniform(Value *V);
+ int Key = Index + SmallestKey;
- unsigned getMaxSafeDepDistBytes() { return MaxSafeDepDistBytes; }
+ // Skip if there is already a member with the same index.
+ if (Members.count(Key))
+ return false;
-private:
- void emitAnalysis(Report &Message) {
- emitLoopAnalysis(Message, TheFunction, TheLoop);
- }
+ if (Key > LargestKey) {
+ // The largest index is always less than the interleave factor.
+ if (Index >= static_cast<int>(Factor))
+ return false;
- /// We need to check that all of the pointers in this list are disjoint
- /// at runtime.
- RuntimePointerCheck PtrRtCheck;
- Function *TheFunction;
- Loop *TheLoop;
- ScalarEvolution *SE;
- const DataLayout *DL;
- const TargetLibraryInfo *TLI;
- AliasAnalysis *AA;
- DominatorTree *DT;
+ LargestKey = Key;
+ } else if (Key < SmallestKey) {
+ // The largest index is always less than the interleave factor.
+ if (LargestKey - Key >= static_cast<int>(Factor))
+ return false;
- unsigned NumLoads;
- unsigned NumStores;
+ SmallestKey = Key;
+ }
- unsigned MaxSafeDepDistBytes;
+ // It's always safe to select the minimum alignment.
+ Align = std::min(Align, NewAlign);
+ Members[Key] = Instr;
+ return true;
+ }
- /// \brief Vectorizer parameters used by the analysis.
- VectorizerParams VectParams;
-};
-} // end anonymous namespace
+ /// \brief Get the member with the given index \p Index
+ ///
+ /// \returns nullptr if contains no such member.
+ Instruction *getMember(unsigned Index) const {
+ int Key = SmallestKey + Index;
+ if (!Members.count(Key))
+ return nullptr;
-/// LoopVectorizationLegality checks if it is legal to vectorize a loop, and
-/// to what vectorization factor.
-/// This class does not look at the profitability of vectorization, only the
-/// legality. This class has two main kinds of checks:
-/// * Memory checks - The code in canVectorizeMemory checks if vectorization
-/// will change the order of memory accesses in a way that will change the
-/// correctness of the program.
-/// * Scalars checks - The code in canVectorizeInstrs and canVectorizeMemory
-/// checks for a number of different conditions, such as the availability of a
-/// single induction variable, that all types are supported and vectorize-able,
-/// etc. This code reflects the capabilities of InnerLoopVectorizer.
-/// This class is also used by InnerLoopVectorizer for identifying
-/// induction variable and the different reduction variables.
-class LoopVectorizationLegality {
-public:
- LoopVectorizationLegality(Loop *L, ScalarEvolution *SE, const DataLayout *DL,
- DominatorTree *DT, TargetLibraryInfo *TLI,
- AliasAnalysis *AA, Function *F,
- const TargetTransformInfo *TTI)
- : NumPredStores(0), TheLoop(L), SE(SE), DL(DL), TLI(TLI), TheFunction(F),
- TTI(TTI), Induction(nullptr), WidestIndTy(nullptr),
- LAA(F, L, SE, DL, TLI, AA, DT,
- {MaxVectorWidth, VectorizationFactor, VectorizationInterleave,
- RuntimeMemoryCheckThreshold}),
- HasFunNoNaNAttr(false) {
- }
-
- /// This enum represents the kinds of reductions that we support.
- enum ReductionKind {
- RK_NoReduction, ///< Not a reduction.
- RK_IntegerAdd, ///< Sum of integers.
- RK_IntegerMult, ///< Product of integers.
- RK_IntegerOr, ///< Bitwise or logical OR of numbers.
- RK_IntegerAnd, ///< Bitwise or logical AND of numbers.
- RK_IntegerXor, ///< Bitwise or logical XOR of numbers.
- RK_IntegerMinMax, ///< Min/max implemented in terms of select(cmp()).
- RK_FloatAdd, ///< Sum of floats.
- RK_FloatMult, ///< Product of floats.
- RK_FloatMinMax ///< Min/max implemented in terms of select(cmp()).
- };
+ return Members.find(Key)->second;
+ }
- /// This enum represents the kinds of inductions that we support.
- enum InductionKind {
- IK_NoInduction, ///< Not an induction variable.
- IK_IntInduction, ///< Integer induction variable. Step = C.
- IK_PtrInduction ///< Pointer induction var. Step = C / sizeof(elem).
- };
+ /// \brief Get the index for the given member. Unlike the key in the member
+ /// map, the index starts from 0.
+ unsigned getIndex(Instruction *Instr) const {
+ for (auto I : Members)
+ if (I.second == Instr)
+ return I.first - SmallestKey;
- // This enum represents the kind of minmax reduction.
- enum MinMaxReductionKind {
- MRK_Invalid,
- MRK_UIntMin,
- MRK_UIntMax,
- MRK_SIntMin,
- MRK_SIntMax,
- MRK_FloatMin,
- MRK_FloatMax
- };
+ llvm_unreachable("InterleaveGroup contains no such member");
+ }
- /// This struct holds information about reduction variables.
- struct ReductionDescriptor {
- ReductionDescriptor() : StartValue(nullptr), LoopExitInstr(nullptr),
- Kind(RK_NoReduction), MinMaxKind(MRK_Invalid) {}
+ Instruction *getInsertPos() const { return InsertPos; }
+ void setInsertPos(Instruction *Inst) { InsertPos = Inst; }
- ReductionDescriptor(Value *Start, Instruction *Exit, ReductionKind K,
- MinMaxReductionKind MK)
- : StartValue(Start), LoopExitInstr(Exit), Kind(K), MinMaxKind(MK) {}
+private:
+ unsigned Factor; // Interleave Factor.
+ bool Reverse;
+ unsigned Align;
+ DenseMap<int, Instruction *> Members;
+ int SmallestKey;
+ int LargestKey;
+
+ // To avoid breaking dependences, vectorized instructions of an interleave
+ // group should be inserted at either the first load or the last store in
+ // program order.
+ //
+ // E.g. %even = load i32 // Insert Position
+ // %add = add i32 %even // Use of %even
+ // %odd = load i32
+ //
+ // store i32 %even
+ // %odd = add i32 // Def of %odd
+ // store i32 %odd // Insert Position
+ Instruction *InsertPos;
+};
- // The starting value of the reduction.
- // It does not have to be zero!
- TrackingVH<Value> StartValue;
- // The instruction who's value is used outside the loop.
- Instruction *LoopExitInstr;
- // The kind of the reduction.
- ReductionKind Kind;
- // If this a min/max reduction the kind of reduction.
- MinMaxReductionKind MinMaxKind;
- };
+/// \brief Drive the analysis of interleaved memory accesses in the loop.
+///
+/// Use this class to analyze interleaved accesses only when we can vectorize
+/// a loop. Otherwise it's meaningless to do analysis as the vectorization
+/// on interleaved accesses is unsafe.
+///
+/// The analysis collects interleave groups and records the relationships
+/// between the member and the group in a map.
+class InterleavedAccessInfo {
+public:
+ InterleavedAccessInfo(ScalarEvolution *SE, Loop *L, DominatorTree *DT)
+ : SE(SE), TheLoop(L), DT(DT) {}
- /// This POD struct holds information about a potential reduction operation.
- struct ReductionInstDesc {
- ReductionInstDesc(bool IsRedux, Instruction *I) :
- IsReduction(IsRedux), PatternLastInst(I), MinMaxKind(MRK_Invalid) {}
-
- ReductionInstDesc(Instruction *I, MinMaxReductionKind K) :
- IsReduction(true), PatternLastInst(I), MinMaxKind(K) {}
-
- // Is this instruction a reduction candidate.
- bool IsReduction;
- // The last instruction in a min/max pattern (select of the select(icmp())
- // pattern), or the current reduction instruction otherwise.
- Instruction *PatternLastInst;
- // If this is a min/max pattern the comparison predicate.
- MinMaxReductionKind MinMaxKind;
- };
+ ~InterleavedAccessInfo() {
+ SmallSet<InterleaveGroup *, 4> DelSet;
+ // Avoid releasing a pointer twice.
+ for (auto &I : InterleaveGroupMap)
+ DelSet.insert(I.second);
+ for (auto *Ptr : DelSet)
+ delete Ptr;
+ }
- /// A struct for saving information about induction variables.
- struct InductionInfo {
- InductionInfo(Value *Start, InductionKind K, ConstantInt *Step)
- : StartValue(Start), IK(K), StepValue(Step) {
- assert(IK != IK_NoInduction && "Not an induction");
- assert(StartValue && "StartValue is null");
- assert(StepValue && !StepValue->isZero() && "StepValue is zero");
- assert((IK != IK_PtrInduction || StartValue->getType()->isPointerTy()) &&
- "StartValue is not a pointer for pointer induction");
- assert((IK != IK_IntInduction || StartValue->getType()->isIntegerTy()) &&
- "StartValue is not an integer for integer induction");
- assert(StepValue->getType()->isIntegerTy() &&
- "StepValue is not an integer");
- }
- InductionInfo()
- : StartValue(nullptr), IK(IK_NoInduction), StepValue(nullptr) {}
+ /// \brief Analyze the interleaved accesses and collect them in interleave
+ /// groups. Substitute symbolic strides using \p Strides.
+ void analyzeInterleaving(const ValueToValueMap &Strides);
- /// Get the consecutive direction. Returns:
- /// 0 - unknown or non-consecutive.
- /// 1 - consecutive and increasing.
- /// -1 - consecutive and decreasing.
- int getConsecutiveDirection() const {
- if (StepValue && (StepValue->isOne() || StepValue->isMinusOne()))
- return StepValue->getSExtValue();
- return 0;
- }
+ /// \brief Check if \p Instr belongs to any interleave group.
+ bool isInterleaved(Instruction *Instr) const {
+ return InterleaveGroupMap.count(Instr);
+ }
- /// Compute the transformed value of Index at offset StartValue using step
- /// StepValue.
- /// For integer induction, returns StartValue + Index * StepValue.
- /// For pointer induction, returns StartValue[Index * StepValue].
- /// FIXME: The newly created binary instructions should contain nsw/nuw
- /// flags, which can be found from the original scalar operations.
- Value *transform(IRBuilder<> &B, Value *Index) const {
- switch (IK) {
- case IK_IntInduction:
- assert(Index->getType() == StartValue->getType() &&
- "Index type does not match StartValue type");
- if (StepValue->isMinusOne())
- return B.CreateSub(StartValue, Index);
- if (!StepValue->isOne())
- Index = B.CreateMul(Index, StepValue);
- return B.CreateAdd(StartValue, Index);
+ /// \brief Get the interleave group that \p Instr belongs to.
+ ///
+ /// \returns nullptr if doesn't have such group.
+ InterleaveGroup *getInterleaveGroup(Instruction *Instr) const {
+ if (InterleaveGroupMap.count(Instr))
+ return InterleaveGroupMap.find(Instr)->second;
+ return nullptr;
+ }
- case IK_PtrInduction:
- if (StepValue->isMinusOne())
- Index = B.CreateNeg(Index);
- else if (!StepValue->isOne())
- Index = B.CreateMul(Index, StepValue);
- return B.CreateGEP(StartValue, Index);
+private:
+ ScalarEvolution *SE;
+ Loop *TheLoop;
+ DominatorTree *DT;
- case IK_NoInduction:
- return nullptr;
- }
- llvm_unreachable("invalid enum");
- }
+ /// Holds the relationships between the members and the interleave group.
+ DenseMap<Instruction *, InterleaveGroup *> InterleaveGroupMap;
- /// Start value.
- TrackingVH<Value> StartValue;
- /// Induction kind.
- InductionKind IK;
- /// Step value.
- ConstantInt *StepValue;
- };
+ /// \brief The descriptor for a strided memory access.
+ struct StrideDescriptor {
+ StrideDescriptor(int Stride, const SCEV *Scev, unsigned Size,
+ unsigned Align)
+ : Stride(Stride), Scev(Scev), Size(Size), Align(Align) {}
- /// ReductionList contains the reduction descriptors for all
- /// of the reductions that were found in the loop.
- typedef DenseMap<PHINode*, ReductionDescriptor> ReductionList;
+ StrideDescriptor() : Stride(0), Scev(nullptr), Size(0), Align(0) {}
- /// InductionList saves induction variables and maps them to the
- /// induction descriptor.
- typedef MapVector<PHINode*, InductionInfo> InductionList;
+ int Stride; // The access's stride. It is negative for a reverse access.
+ const SCEV *Scev; // The scalar expression of this access
+ unsigned Size; // The size of the memory object.
+ unsigned Align; // The alignment of this access.
+ };
- /// Returns true if it is legal to vectorize this loop.
- /// This does not mean that it is profitable to vectorize this
- /// loop, only that it is legal to do so.
- bool canVectorize();
+ /// \brief Create a new interleave group with the given instruction \p Instr,
+ /// stride \p Stride and alignment \p Align.
+ ///
+ /// \returns the newly created interleave group.
+ InterleaveGroup *createInterleaveGroup(Instruction *Instr, int Stride,
+ unsigned Align) {
+ assert(!InterleaveGroupMap.count(Instr) &&
+ "Already in an interleaved access group");
+ InterleaveGroupMap[Instr] = new InterleaveGroup(Instr, Stride, Align);
+ return InterleaveGroupMap[Instr];
+ }
- /// Returns the Induction variable.
- PHINode *getInduction() { return Induction; }
+ /// \brief Release the group and remove all the relationships.
+ void releaseGroup(InterleaveGroup *Group) {
+ for (unsigned i = 0; i < Group->getFactor(); i++)
+ if (Instruction *Member = Group->getMember(i))
+ InterleaveGroupMap.erase(Member);
- /// Returns the reduction variables found in the loop.
- ReductionList *getReductionVars() { return &Reductions; }
+ delete Group;
+ }
- /// Returns the induction variables found in the loop.
- InductionList *getInductionVars() { return &Inductions; }
+ /// \brief Collect all the accesses with a constant stride in program order.
+ void collectConstStridedAccesses(
+ MapVector<Instruction *, StrideDescriptor> &StrideAccesses,
+ const ValueToValueMap &Strides);
+};
- /// Returns the widest induction type.
- Type *getWidestInductionType() { return WidestIndTy; }
+/// Utility class for getting and setting loop vectorizer hints in the form
+/// of loop metadata.
+/// This class keeps a number of loop annotations locally (as member variables)
+/// and can, upon request, write them back as metadata on the loop. It will
+/// initially scan the loop for existing metadata, and will update the local
+/// values based on information in the loop.
+/// We cannot write all values to metadata, as the mere presence of some info,
+/// for example 'force', means a decision has been made. So, we need to be
+/// careful NOT to add them if the user hasn't specifically asked so.
+class LoopVectorizeHints {
+ enum HintKind {
+ HK_WIDTH,
+ HK_UNROLL,
+ HK_FORCE
+ };
- /// Returns True if V is an induction variable in this loop.
- bool isInductionVariable(const Value *V);
+ /// Hint - associates name and validation with the hint value.
+ struct Hint {
+ const char * Name;
+ unsigned Value; // This may have to change for non-numeric values.
+ HintKind Kind;
- /// Return true if the block BB needs to be predicated in order for the loop
- /// to be vectorized.
- bool blockNeedsPredication(BasicBlock *BB);
+ Hint(const char * Name, unsigned Value, HintKind Kind)
+ : Name(Name), Value(Value), Kind(Kind) { }
- /// Check if this pointer is consecutive when vectorizing. This happens
- /// when the last index of the GEP is the induction variable, or that the
- /// pointer itself is an induction variable.
- /// This check allows us to vectorize A[idx] into a wide load/store.
- /// Returns:
- /// 0 - Stride is unknown or non-consecutive.
- /// 1 - Address is consecutive.
- /// -1 - Address is consecutive, and decreasing.
- int isConsecutivePtr(Value *Ptr);
+ bool validate(unsigned Val) {
+ switch (Kind) {
+ case HK_WIDTH:
+ return isPowerOf2_32(Val) && Val <= VectorizerParams::MaxVectorWidth;
+ case HK_UNROLL:
+ return isPowerOf2_32(Val) && Val <= MaxInterleaveFactor;
+ case HK_FORCE:
+ return (Val <= 1);
+ }
+ return false;
+ }
+ };
- /// Returns true if the value V is uniform within the loop.
- bool isUniform(Value *V);
+ /// Vectorization width.
+ Hint Width;
+ /// Vectorization interleave factor.
+ Hint Interleave;
+ /// Vectorization forced
+ Hint Force;
- /// Returns true if this instruction will remain scalar after vectorization.
- bool isUniformAfterVectorization(Instruction* I) { return Uniforms.count(I); }
+ /// Return the loop metadata prefix.
+ static StringRef Prefix() { return "llvm.loop."; }
- /// Returns the information that we collected about runtime memory check.
- RuntimePointerCheck *getRuntimePointerCheck() {
- return LAA.getRuntimePointerCheck();
- }
+public:
+ enum ForceKind {
+ FK_Undefined = -1, ///< Not selected.
+ FK_Disabled = 0, ///< Forcing disabled.
+ FK_Enabled = 1, ///< Forcing enabled.
+ };
- /// This function returns the identity element (or neutral element) for
- /// the operation K.
- static Constant *getReductionIdentity(ReductionKind K, Type *Tp);
+ LoopVectorizeHints(const Loop *L, bool DisableInterleaving)
+ : Width("vectorize.width", VectorizerParams::VectorizationFactor,
+ HK_WIDTH),
+ Interleave("interleave.count", DisableInterleaving, HK_UNROLL),
+ Force("vectorize.enable", FK_Undefined, HK_FORCE),
+ TheLoop(L) {
+ // Populate values with existing loop metadata.
+ getHintsFromMetadata();
- unsigned getMaxSafeDepDistBytes() { return LAA.getMaxSafeDepDistBytes(); }
+ // force-vector-interleave overrides DisableInterleaving.
+ if (VectorizerParams::isInterleaveForced())
+ Interleave.Value = VectorizerParams::VectorizationInterleave;
- bool hasStride(Value *V) { return StrideSet.count(V); }
- bool mustCheckStrides() { return !StrideSet.empty(); }
- SmallPtrSet<Value *, 8>::iterator strides_begin() {
- return StrideSet.begin();
+ DEBUG(if (DisableInterleaving && Interleave.Value == 1) dbgs()
+ << "LV: Interleaving disabled by the pass manager\n");
}
- SmallPtrSet<Value *, 8>::iterator strides_end() { return StrideSet.end(); }
- /// Returns true if the target machine supports masked store operation
- /// for the given \p DataType and kind of access to \p Ptr.
- bool isLegalMaskedStore(Type *DataType, Value *Ptr) {
- return TTI->isLegalMaskedStore(DataType, isConsecutivePtr(Ptr));
- }
- /// Returns true if the target machine supports masked load operation
- /// for the given \p DataType and kind of access to \p Ptr.
- bool isLegalMaskedLoad(Type *DataType, Value *Ptr) {
- return TTI->isLegalMaskedLoad(DataType, isConsecutivePtr(Ptr));
- }
- /// Returns true if vector representation of the instruction \p I
- /// requires mask.
- bool isMaskRequired(const Instruction* I) {
- return (MaskedOp.count(I) != 0);
- }
- unsigned getNumStores() const {
- return NumStores;
- }
- unsigned getNumLoads() const {
- return NumLoads;
- }
- unsigned getNumPredStores() const {
- return NumPredStores;
+ /// Mark the loop L as already vectorized by setting the width to 1.
+ void setAlreadyVectorized() {
+ Width.Value = Interleave.Value = 1;
+ Hint Hints[] = {Width, Interleave};
+ writeHintsToMetadata(Hints);
}
-private:
- /// Check if a single basic block loop is vectorizable.
- /// At this point we know that this is a loop with a constant trip count
- /// and we only need to check individual instructions.
- bool canVectorizeInstrs();
-
- /// When we vectorize loops we may change the order in which
- /// we read and write from memory. This method checks if it is
- /// legal to vectorize the code, considering only memory constrains.
- /// Returns true if the loop is vectorizable
- bool canVectorizeMemory();
- /// Return true if we can vectorize this loop using the IF-conversion
- /// transformation.
- bool canVectorizeWithIfConvert();
+ bool allowVectorization(Function *F, Loop *L, bool AlwaysVectorize) const {
+ if (getForce() == LoopVectorizeHints::FK_Disabled) {
+ DEBUG(dbgs() << "LV: Not vectorizing: #pragma vectorize disable.\n");
+ emitOptimizationRemarkAnalysis(F->getContext(), DEBUG_TYPE, *F,
+ L->getStartLoc(), emitRemark());
+ return false;
+ }
- /// Collect the variables that need to stay uniform after vectorization.
- void collectLoopUniforms();
+ if (!AlwaysVectorize && getForce() != LoopVectorizeHints::FK_Enabled) {
+ DEBUG(dbgs() << "LV: Not vectorizing: No #pragma vectorize enable.\n");
+ emitOptimizationRemarkAnalysis(F->getContext(), DEBUG_TYPE, *F,
+ L->getStartLoc(), emitRemark());
+ return false;
+ }
- /// Return true if all of the instructions in the block can be speculatively
- /// executed. \p SafePtrs is a list of addresses that are known to be legal
- /// and we know that we can read from them without segfault.
- bool blockCanBePredicated(BasicBlock *BB, SmallPtrSetImpl<Value *> &SafePtrs);
+ if (getWidth() == 1 && getInterleave() == 1) {
+ // FIXME: Add a separate metadata to indicate when the loop has already
+ // been vectorized instead of setting width and count to 1.
+ DEBUG(dbgs() << "LV: Not vectorizing: Disabled/already vectorized.\n");
+ // FIXME: Add interleave.disable metadata. This will allow
+ // vectorize.disable to be used without disabling the pass and errors
+ // to differentiate between disabled vectorization and a width of 1.
+ emitOptimizationRemarkAnalysis(
+ F->getContext(), DEBUG_TYPE, *F, L->getStartLoc(),
+ "loop not vectorized: vectorization and interleaving are explicitly "
+ "disabled, or vectorize width and interleave count are both set to "
+ "1");
+ return false;
+ }
- /// Returns True, if 'Phi' is the kind of reduction variable for type
- /// 'Kind'. If this is a reduction variable, it adds it to ReductionList.
- bool AddReductionVar(PHINode *Phi, ReductionKind Kind);
- /// Returns a struct describing if the instruction 'I' can be a reduction
- /// variable of type 'Kind'. If the reduction is a min/max pattern of
- /// select(icmp()) this function advances the instruction pointer 'I' from the
- /// compare instruction to the select instruction and stores this pointer in
- /// 'PatternLastInst' member of the returned struct.
- ReductionInstDesc isReductionInstr(Instruction *I, ReductionKind Kind,
- ReductionInstDesc &Desc);
- /// Returns true if the instruction is a Select(ICmp(X, Y), X, Y) instruction
- /// pattern corresponding to a min(X, Y) or max(X, Y).
- static ReductionInstDesc isMinMaxSelectCmpPattern(Instruction *I,
- ReductionInstDesc &Prev);
- /// Returns the induction kind of Phi and record the step. This function may
- /// return NoInduction if the PHI is not an induction variable.
- InductionKind isInductionVariable(PHINode *Phi, ConstantInt *&StepValue);
+ return true;
+ }
- /// \brief Collect memory access with loop invariant strides.
- ///
- /// Looks for accesses like "a[i * StrideA]" where "StrideA" is loop
- /// invariant.
- void collectStridedAccess(Value *LoadOrStoreInst);
+ /// Dumps all the hint information.
+ std::string emitRemark() const {
+ VectorizationReport R;
+ if (Force.Value == LoopVectorizeHints::FK_Disabled)
+ R << "vectorization is explicitly disabled";
+ else {
+ R << "use -Rpass-analysis=loop-vectorize for more info";
+ if (Force.Value == LoopVectorizeHints::FK_Enabled) {
+ R << " (Force=true";
+ if (Width.Value != 0)
+ R << ", Vector Width=" << Width.Value;
+ if (Interleave.Value != 0)
+ R << ", Interleave Count=" << Interleave.Value;
+ R << ")";
+ }
+ }
- /// Report an analysis message to assist the user in diagnosing loops that are
- /// not vectorized.
- void emitAnalysis(VectorizationReport &Message) {
- VectorizationReport::emitAnalysis(Message, TheFunction, TheLoop);
+ return R.str();
}
- unsigned NumLoads;
- unsigned NumStores;
- unsigned NumPredStores;
+ unsigned getWidth() const { return Width.Value; }
+ unsigned getInterleave() const { return Interleave.Value; }
+ enum ForceKind getForce() const { return (ForceKind)Force.Value; }
+ bool isForced() const {
+ return getForce() == LoopVectorizeHints::FK_Enabled || getWidth() > 1 ||
+ getInterleave() > 1;
+ }
- /// The loop that we evaluate.
- Loop *TheLoop;
- /// Scev analysis.
- ScalarEvolution *SE;
- /// DataLayout analysis.
- const DataLayout *DL;
- /// Target Library Info.
- TargetLibraryInfo *TLI;
- /// Parent function
- Function *TheFunction;
- /// Target Transform Info
- const TargetTransformInfo *TTI;
+private:
+ /// Find hints specified in the loop metadata and update local values.
+ void getHintsFromMetadata() {
+ MDNode *LoopID = TheLoop->getLoopID();
+ if (!LoopID)
+ return;
- // --- vectorization state --- //
+ // First operand should refer to the loop id itself.
+ assert(LoopID->getNumOperands() > 0 && "requires at least one operand");
+ assert(LoopID->getOperand(0) == LoopID && "invalid loop id");
- /// Holds the integer induction variable. This is the counter of the
- /// loop.
- PHINode *Induction;
- /// Holds the reduction variables.
- ReductionList Reductions;
- /// Holds all of the induction variables that we found in the loop.
- /// Notice that inductions don't need to start at zero and that induction
- /// variables can be pointers.
- InductionList Inductions;
- /// Holds the widest induction type encountered.
- Type *WidestIndTy;
+ for (unsigned i = 1, ie = LoopID->getNumOperands(); i < ie; ++i) {
+ const MDString *S = nullptr;
+ SmallVector<Metadata *, 4> Args;
- /// Allowed outside users. This holds the reduction
- /// vars which can be accessed from outside the loop.
- SmallPtrSet<Value*, 4> AllowedExit;
- /// This set holds the variables which are known to be uniform after
- /// vectorization.
- SmallPtrSet<Instruction*, 4> Uniforms;
- LoopAccessAnalysis LAA;
- /// Can we assume the absence of NaNs.
- bool HasFunNoNaNAttr;
+ // The expected hint is either a MDString or a MDNode with the first
+ // operand a MDString.
+ if (const MDNode *MD = dyn_cast<MDNode>(LoopID->getOperand(i))) {
+ if (!MD || MD->getNumOperands() == 0)
+ continue;
+ S = dyn_cast<MDString>(MD->getOperand(0));
+ for (unsigned i = 1, ie = MD->getNumOperands(); i < ie; ++i)
+ Args.push_back(MD->getOperand(i));
+ } else {
+ S = dyn_cast<MDString>(LoopID->getOperand(i));
+ assert(Args.size() == 0 && "too many arguments for MDString");
+ }
- ValueToValueMap Strides;
- SmallPtrSet<Value *, 8> StrideSet;
-
- /// While vectorizing these instructions we have to generate a
- /// call to the appropriate masked intrinsic
- SmallPtrSet<const Instruction*, 8> MaskedOp;
-};
+ if (!S)
+ continue;
-/// LoopVectorizationCostModel - estimates the expected speedups due to
-/// vectorization.
-/// In many cases vectorization is not profitable. This can happen because of
-/// a number of reasons. In this class we mainly attempt to predict the
-/// expected speedup/slowdowns due to the supported instruction set. We use the
-/// TargetTransformInfo to query the different backends for the cost of
-/// different operations.
-class LoopVectorizationCostModel {
-public:
- LoopVectorizationCostModel(Loop *L, ScalarEvolution *SE, LoopInfo *LI,
- LoopVectorizationLegality *Legal,
- const TargetTransformInfo &TTI,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
- AssumptionCache *AC, const Function *F,
- const LoopVectorizeHints *Hints)
- : TheLoop(L), SE(SE), LI(LI), Legal(Legal), TTI(TTI), DL(DL), TLI(TLI),
- TheFunction(F), Hints(Hints) {
- CodeMetrics::collectEphemeralValues(L, AC, EphValues);
+ // Check if the hint starts with the loop metadata prefix.
+ StringRef Name = S->getString();
+ if (Args.size() == 1)
+ setHint(Name, Args[0]);
+ }
}
- /// Information about vectorization costs
- struct VectorizationFactor {
- unsigned Width; // Vector width with best cost
- unsigned Cost; // Cost of the loop with that width
- };
- /// \return The most profitable vectorization factor and the cost of that VF.
- /// This method checks every power of two up to VF. If UserVF is not ZERO
- /// then this vectorization factor will be selected if vectorization is
- /// possible.
- VectorizationFactor selectVectorizationFactor(bool OptForSize);
-
- /// \return The size (in bits) of the widest type in the code that
- /// needs to be vectorized. We ignore values that remain scalar such as
- /// 64 bit loop indices.
- unsigned getWidestType();
-
- /// \return The most profitable unroll factor.
- /// If UserUF is non-zero then this method finds the best unroll-factor
- /// based on register pressure and other parameters.
- /// VF and LoopCost are the selected vectorization factor and the cost of the
- /// selected VF.
- unsigned selectUnrollFactor(bool OptForSize, unsigned VF, unsigned LoopCost);
+ /// Checks string hint with one operand and set value if valid.
+ void setHint(StringRef Name, Metadata *Arg) {
+ if (!Name.startswith(Prefix()))
+ return;
+ Name = Name.substr(Prefix().size(), StringRef::npos);
- /// \brief A struct that represents some properties of the register usage
- /// of a loop.
- struct RegisterUsage {
- /// Holds the number of loop invariant values that are used in the loop.
- unsigned LoopInvariantRegs;
- /// Holds the maximum number of concurrent live intervals in the loop.
- unsigned MaxLocalUsers;
- /// Holds the number of instructions in the loop.
- unsigned NumInstructions;
- };
+ const ConstantInt *C = mdconst::dyn_extract<ConstantInt>(Arg);
+ if (!C) return;
+ unsigned Val = C->getZExtValue();
- /// \return information about the register usage of the loop.
- RegisterUsage calculateRegisterUsage();
+ Hint *Hints[] = {&Width, &Interleave, &Force};
+ for (auto H : Hints) {
+ if (Name == H->Name) {
+ if (H->validate(Val))
+ H->Value = Val;
+ else
+ DEBUG(dbgs() << "LV: ignoring invalid hint '" << Name << "'\n");
+ break;
+ }
+ }
+ }
-private:
- /// Returns the expected execution cost. The unit of the cost does
- /// not matter because we use the 'cost' units to compare different
- /// vector widths. The cost that is returned is *not* normalized by
- /// the factor width.
- unsigned expectedCost(unsigned VF);
+ /// Create a new hint from name / value pair.
+ MDNode *createHintMetadata(StringRef Name, unsigned V) const {
+ LLVMContext &Context = TheLoop->getHeader()->getContext();
+ Metadata *MDs[] = {MDString::get(Context, Name),
+ ConstantAsMetadata::get(
+ ConstantInt::get(Type::getInt32Ty(Context), V))};
+ return MDNode::get(Context, MDs);
+ }
- /// Returns the execution time cost of an instruction for a given vector
- /// width. Vector width of one means scalar.
- unsigned getInstructionCost(Instruction *I, unsigned VF);
+ /// Matches metadata with hint name.
+ bool matchesHintMetadataName(MDNode *Node, ArrayRef<Hint> HintTypes) {
+ MDString* Name = dyn_cast<MDString>(Node->getOperand(0));
+ if (!Name)
+ return false;
- /// A helper function for converting Scalar types to vector types.
- /// If the incoming type is void, we return void. If the VF is 1, we return
- /// the scalar type.
- static Type* ToVectorTy(Type *Scalar, unsigned VF);
-
- /// Returns whether the instruction is a load or store and will be a emitted
- /// as a vector operation.
- bool isConsecutiveLoadOrStore(Instruction *I);
-
- /// Report an analysis message to assist the user in diagnosing loops that are
- /// not vectorized.
- void emitAnalysis(VectorizationReport &Message) {
- VectorizationReport::emitAnalysis(Message, TheFunction, TheLoop);
- }
-
- /// Values used only by @llvm.assume calls.
- SmallPtrSet<const Value *, 32> EphValues;
-
- /// The loop that we evaluate.
- Loop *TheLoop;
- /// Scev analysis.
- ScalarEvolution *SE;
- /// Loop Info analysis.
- LoopInfo *LI;
- /// Vectorization legality.
- LoopVectorizationLegality *Legal;
- /// Vector target information.
- const TargetTransformInfo &TTI;
- /// Target data layout information.
- const DataLayout *DL;
- /// Target Library Info.
- const TargetLibraryInfo *TLI;
- const Function *TheFunction;
- // Loop Vectorize Hint.
- const LoopVectorizeHints *Hints;
-};
-
-/// Utility class for getting and setting loop vectorizer hints in the form
-/// of loop metadata.
-/// This class keeps a number of loop annotations locally (as member variables)
-/// and can, upon request, write them back as metadata on the loop. It will
-/// initially scan the loop for existing metadata, and will update the local
-/// values based on information in the loop.
-/// We cannot write all values to metadata, as the mere presence of some info,
-/// for example 'force', means a decision has been made. So, we need to be
-/// careful NOT to add them if the user hasn't specifically asked so.
-class LoopVectorizeHints {
- enum HintKind {
- HK_WIDTH,
- HK_UNROLL,
- HK_FORCE
- };
-
- /// Hint - associates name and validation with the hint value.
- struct Hint {
- const char * Name;
- unsigned Value; // This may have to change for non-numeric values.
- HintKind Kind;
-
- Hint(const char * Name, unsigned Value, HintKind Kind)
- : Name(Name), Value(Value), Kind(Kind) { }
-
- bool validate(unsigned Val) {
- switch (Kind) {
- case HK_WIDTH:
- return isPowerOf2_32(Val) && Val <= MaxVectorWidth;
- case HK_UNROLL:
- return isPowerOf2_32(Val) && Val <= MaxInterleaveFactor;
- case HK_FORCE:
- return (Val <= 1);
- }
- return false;
- }
- };
-
- /// Vectorization width.
- Hint Width;
- /// Vectorization interleave factor.
- Hint Interleave;
- /// Vectorization forced
- Hint Force;
-
- /// Return the loop metadata prefix.
- static StringRef Prefix() { return "llvm.loop."; }
-
-public:
- enum ForceKind {
- FK_Undefined = -1, ///< Not selected.
- FK_Disabled = 0, ///< Forcing disabled.
- FK_Enabled = 1, ///< Forcing enabled.
- };
-
- LoopVectorizeHints(const Loop *L, bool DisableInterleaving)
- : Width("vectorize.width", VectorizationFactor, HK_WIDTH),
- Interleave("interleave.count", DisableInterleaving, HK_UNROLL),
- Force("vectorize.enable", FK_Undefined, HK_FORCE),
- TheLoop(L) {
- // Populate values with existing loop metadata.
- getHintsFromMetadata();
-
- // force-vector-interleave overrides DisableInterleaving.
- if (VectorizationInterleave.getNumOccurrences() > 0)
- Interleave.Value = VectorizationInterleave;
-
- DEBUG(if (DisableInterleaving && Interleave.Value == 1) dbgs()
- << "LV: Interleaving disabled by the pass manager\n");
- }
-
- /// Mark the loop L as already vectorized by setting the width to 1.
- void setAlreadyVectorized() {
- Width.Value = Interleave.Value = 1;
- Hint Hints[] = {Width, Interleave};
- writeHintsToMetadata(Hints);
- }
-
- /// Dumps all the hint information.
- std::string emitRemark() const {
- VectorizationReport R;
- if (Force.Value == LoopVectorizeHints::FK_Disabled)
- R << "vectorization is explicitly disabled";
- else {
- R << "use -Rpass-analysis=loop-vectorize for more info";
- if (Force.Value == LoopVectorizeHints::FK_Enabled) {
- R << " (Force=true";
- if (Width.Value != 0)
- R << ", Vector Width=" << Width.Value;
- if (Interleave.Value != 0)
- R << ", Interleave Count=" << Interleave.Value;
- R << ")";
- }
- }
-
- return R.str();
- }
-
- unsigned getWidth() const { return Width.Value; }
- unsigned getInterleave() const { return Interleave.Value; }
- enum ForceKind getForce() const { return (ForceKind)Force.Value; }
-
-private:
- /// Find hints specified in the loop metadata and update local values.
- void getHintsFromMetadata() {
- MDNode *LoopID = TheLoop->getLoopID();
- if (!LoopID)
- return;
-
- // First operand should refer to the loop id itself.
- assert(LoopID->getNumOperands() > 0 && "requires at least one operand");
- assert(LoopID->getOperand(0) == LoopID && "invalid loop id");
-
- for (unsigned i = 1, ie = LoopID->getNumOperands(); i < ie; ++i) {
- const MDString *S = nullptr;
- SmallVector<Metadata *, 4> Args;
-
- // The expected hint is either a MDString or a MDNode with the first
- // operand a MDString.
- if (const MDNode *MD = dyn_cast<MDNode>(LoopID->getOperand(i))) {
- if (!MD || MD->getNumOperands() == 0)
- continue;
- S = dyn_cast<MDString>(MD->getOperand(0));
- for (unsigned i = 1, ie = MD->getNumOperands(); i < ie; ++i)
- Args.push_back(MD->getOperand(i));
- } else {
- S = dyn_cast<MDString>(LoopID->getOperand(i));
- assert(Args.size() == 0 && "too many arguments for MDString");
- }
-
- if (!S)
- continue;
-
- // Check if the hint starts with the loop metadata prefix.
- StringRef Name = S->getString();
- if (Args.size() == 1)
- setHint(Name, Args[0]);
- }
- }
-
- /// Checks string hint with one operand and set value if valid.
- void setHint(StringRef Name, Metadata *Arg) {
- if (!Name.startswith(Prefix()))
- return;
- Name = Name.substr(Prefix().size(), StringRef::npos);
-
- const ConstantInt *C = mdconst::dyn_extract<ConstantInt>(Arg);
- if (!C) return;
- unsigned Val = C->getZExtValue();
-
- Hint *Hints[] = {&Width, &Interleave, &Force};
- for (auto H : Hints) {
- if (Name == H->Name) {
- if (H->validate(Val))
- H->Value = Val;
- else
- DEBUG(dbgs() << "LV: ignoring invalid hint '" << Name << "'\n");
- break;
- }
- }
- }
-
- /// Create a new hint from name / value pair.
- MDNode *createHintMetadata(StringRef Name, unsigned V) const {
- LLVMContext &Context = TheLoop->getHeader()->getContext();
- Metadata *MDs[] = {MDString::get(Context, Name),
- ConstantAsMetadata::get(
- ConstantInt::get(Type::getInt32Ty(Context), V))};
- return MDNode::get(Context, MDs);
- }
-
- /// Matches metadata with hint name.
- bool matchesHintMetadataName(MDNode *Node, ArrayRef<Hint> HintTypes) {
- MDString* Name = dyn_cast<MDString>(Node->getOperand(0));
- if (!Name)
- return false;
-
- for (auto H : HintTypes)
- if (Name->getString().endswith(H.Name))
- return true;
- return false;
- }
+ for (auto H : HintTypes)
+ if (Name->getString().endswith(H.Name))
+ return true;
+ return false;
+ }
/// Sets current hints into loop metadata, keeping other values intact.
void writeHintsToMetadata(ArrayRef<Hint> HintTypes) {
const Loop *TheLoop;
};
+static void emitAnalysisDiag(const Function *TheFunction, const Loop *TheLoop,
+ const LoopVectorizeHints &Hints,
+ const LoopAccessReport &Message) {
+ // If a loop hint is provided the diagnostic is always produced.
+ const char *Name = Hints.isForced() ? DiagnosticInfo::AlwaysPrint : LV_NAME;
+ LoopAccessReport::emitAnalysis(Message, TheFunction, TheLoop, Name);
+}
+
static void emitMissedWarning(Function *F, Loop *L,
const LoopVectorizeHints &LH) {
emitOptimizationRemarkMissed(F->getContext(), DEBUG_TYPE, *F,
}
}
-static void addInnerLoop(Loop &L, SmallVectorImpl<Loop *> &V) {
- if (L.empty())
- return V.push_back(&L);
+/// LoopVectorizationLegality checks if it is legal to vectorize a loop, and
+/// to what vectorization factor.
+/// This class does not look at the profitability of vectorization, only the
+/// legality. This class has two main kinds of checks:
+/// * Memory checks - The code in canVectorizeMemory checks if vectorization
+/// will change the order of memory accesses in a way that will change the
+/// correctness of the program.
+/// * Scalars checks - The code in canVectorizeInstrs and canVectorizeMemory
+/// checks for a number of different conditions, such as the availability of a
+/// single induction variable, that all types are supported and vectorize-able,
+/// etc. This code reflects the capabilities of InnerLoopVectorizer.
+/// This class is also used by InnerLoopVectorizer for identifying
+/// induction variable and the different reduction variables.
+class LoopVectorizationLegality {
+public:
+ LoopVectorizationLegality(Loop *L, ScalarEvolution *SE, DominatorTree *DT,
+ TargetLibraryInfo *TLI, AliasAnalysis *AA,
+ Function *F, const TargetTransformInfo *TTI,
+ LoopAccessAnalysis *LAA,
+ LoopVectorizationRequirements *R,
+ const LoopVectorizeHints *H)
+ : NumPredStores(0), TheLoop(L), SE(SE), TLI(TLI), TheFunction(F),
+ TTI(TTI), DT(DT), LAA(LAA), LAI(nullptr), InterleaveInfo(SE, L, DT),
+ Induction(nullptr), WidestIndTy(nullptr), HasFunNoNaNAttr(false),
+ Requirements(R), Hints(H) {}
- for (Loop *InnerL : L)
- addInnerLoop(*InnerL, V);
-}
+ /// This enum represents the kinds of inductions that we support.
+ enum InductionKind {
+ IK_NoInduction, ///< Not an induction variable.
+ IK_IntInduction, ///< Integer induction variable. Step = C.
+ IK_PtrInduction ///< Pointer induction var. Step = C / sizeof(elem).
+ };
-/// The LoopVectorize Pass.
-struct LoopVectorize : public FunctionPass {
- /// Pass identification, replacement for typeid
- static char ID;
+ /// A struct for saving information about induction variables.
+ struct InductionInfo {
+ InductionInfo(Value *Start, InductionKind K, ConstantInt *Step)
+ : StartValue(Start), IK(K), StepValue(Step) {
+ assert(IK != IK_NoInduction && "Not an induction");
+ assert(StartValue && "StartValue is null");
+ assert(StepValue && !StepValue->isZero() && "StepValue is zero");
+ assert((IK != IK_PtrInduction || StartValue->getType()->isPointerTy()) &&
+ "StartValue is not a pointer for pointer induction");
+ assert((IK != IK_IntInduction || StartValue->getType()->isIntegerTy()) &&
+ "StartValue is not an integer for integer induction");
+ assert(StepValue->getType()->isIntegerTy() &&
+ "StepValue is not an integer");
+ }
+ InductionInfo()
+ : StartValue(nullptr), IK(IK_NoInduction), StepValue(nullptr) {}
- explicit LoopVectorize(bool NoUnrolling = false, bool AlwaysVectorize = true)
- : FunctionPass(ID),
- DisableUnrolling(NoUnrolling),
- AlwaysVectorize(AlwaysVectorize) {
- initializeLoopVectorizePass(*PassRegistry::getPassRegistry());
- }
-
- ScalarEvolution *SE;
- const DataLayout *DL;
- LoopInfo *LI;
- TargetTransformInfo *TTI;
- DominatorTree *DT;
- BlockFrequencyInfo *BFI;
- TargetLibraryInfo *TLI;
- AliasAnalysis *AA;
- AssumptionCache *AC;
- bool DisableUnrolling;
- bool AlwaysVectorize;
-
- BlockFrequency ColdEntryFreq;
-
- bool runOnFunction(Function &F) override {
- SE = &getAnalysis<ScalarEvolution>();
- DataLayoutPass *DLP = getAnalysisIfAvailable<DataLayoutPass>();
- DL = DLP ? &DLP->getDataLayout() : nullptr;
- LI = &getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
- TTI = &getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F);
- DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
- BFI = &getAnalysis<BlockFrequencyInfo>();
- auto *TLIP = getAnalysisIfAvailable<TargetLibraryInfoWrapperPass>();
- TLI = TLIP ? &TLIP->getTLI() : nullptr;
- AA = &getAnalysis<AliasAnalysis>();
- AC = &getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
+ /// Get the consecutive direction. Returns:
+ /// 0 - unknown or non-consecutive.
+ /// 1 - consecutive and increasing.
+ /// -1 - consecutive and decreasing.
+ int getConsecutiveDirection() const {
+ if (StepValue && (StepValue->isOne() || StepValue->isMinusOne()))
+ return StepValue->getSExtValue();
+ return 0;
+ }
- // Compute some weights outside of the loop over the loops. Compute this
- // using a BranchProbability to re-use its scaling math.
- const BranchProbability ColdProb(1, 5); // 20%
- ColdEntryFreq = BlockFrequency(BFI->getEntryFreq()) * ColdProb;
+ /// Compute the transformed value of Index at offset StartValue using step
+ /// StepValue.
+ /// For integer induction, returns StartValue + Index * StepValue.
+ /// For pointer induction, returns StartValue[Index * StepValue].
+ /// FIXME: The newly created binary instructions should contain nsw/nuw
+ /// flags, which can be found from the original scalar operations.
+ Value *transform(IRBuilder<> &B, Value *Index) const {
+ switch (IK) {
+ case IK_IntInduction:
+ assert(Index->getType() == StartValue->getType() &&
+ "Index type does not match StartValue type");
+ if (StepValue->isMinusOne())
+ return B.CreateSub(StartValue, Index);
+ if (!StepValue->isOne())
+ Index = B.CreateMul(Index, StepValue);
+ return B.CreateAdd(StartValue, Index);
- // If the target claims to have no vector registers don't attempt
- // vectorization.
- if (!TTI->getNumberOfRegisters(true))
- return false;
+ case IK_PtrInduction:
+ assert(Index->getType() == StepValue->getType() &&
+ "Index type does not match StepValue type");
+ if (StepValue->isMinusOne())
+ Index = B.CreateNeg(Index);
+ else if (!StepValue->isOne())
+ Index = B.CreateMul(Index, StepValue);
+ return B.CreateGEP(nullptr, StartValue, Index);
- if (!DL) {
- DEBUG(dbgs() << "\nLV: Not vectorizing " << F.getName()
- << ": Missing data layout\n");
- return false;
+ case IK_NoInduction:
+ return nullptr;
+ }
+ llvm_unreachable("invalid enum");
}
- // Build up a worklist of inner-loops to vectorize. This is necessary as
- // the act of vectorizing or partially unrolling a loop creates new loops
- // and can invalidate iterators across the loops.
- SmallVector<Loop *, 8> Worklist;
+ /// Start value.
+ TrackingVH<Value> StartValue;
+ /// Induction kind.
+ InductionKind IK;
+ /// Step value.
+ ConstantInt *StepValue;
+ };
- for (Loop *L : *LI)
- addInnerLoop(*L, Worklist);
+ /// ReductionList contains the reduction descriptors for all
+ /// of the reductions that were found in the loop.
+ typedef DenseMap<PHINode *, RecurrenceDescriptor> ReductionList;
- LoopsAnalyzed += Worklist.size();
+ /// InductionList saves induction variables and maps them to the
+ /// induction descriptor.
+ typedef MapVector<PHINode*, InductionInfo> InductionList;
- // Now walk the identified inner loops.
- bool Changed = false;
- while (!Worklist.empty())
- Changed |= processLoop(Worklist.pop_back_val());
+ /// Returns true if it is legal to vectorize this loop.
+ /// This does not mean that it is profitable to vectorize this
+ /// loop, only that it is legal to do so.
+ bool canVectorize();
- // Process each loop nest in the function.
- return Changed;
- }
+ /// Returns the Induction variable.
+ PHINode *getInduction() { return Induction; }
- bool processLoop(Loop *L) {
- assert(L->empty() && "Only process inner loops.");
+ /// Returns the reduction variables found in the loop.
+ ReductionList *getReductionVars() { return &Reductions; }
-#ifndef NDEBUG
- const std::string DebugLocStr = getDebugLocString(L);
-#endif /* NDEBUG */
+ /// Returns the induction variables found in the loop.
+ InductionList *getInductionVars() { return &Inductions; }
- DEBUG(dbgs() << "\nLV: Checking a loop in \""
- << L->getHeader()->getParent()->getName() << "\" from "
- << DebugLocStr << "\n");
+ /// Returns the widest induction type.
+ Type *getWidestInductionType() { return WidestIndTy; }
- LoopVectorizeHints Hints(L, DisableUnrolling);
+ /// Returns True if V is an induction variable in this loop.
+ bool isInductionVariable(const Value *V);
- DEBUG(dbgs() << "LV: Loop hints:"
- << " force="
- << (Hints.getForce() == LoopVectorizeHints::FK_Disabled
- ? "disabled"
- : (Hints.getForce() == LoopVectorizeHints::FK_Enabled
- ? "enabled"
- : "?")) << " width=" << Hints.getWidth()
- << " unroll=" << Hints.getInterleave() << "\n");
+ /// Return true if the block BB needs to be predicated in order for the loop
+ /// to be vectorized.
+ bool blockNeedsPredication(BasicBlock *BB);
- // Function containing loop
- Function *F = L->getHeader()->getParent();
+ /// Check if this pointer is consecutive when vectorizing. This happens
+ /// when the last index of the GEP is the induction variable, or that the
+ /// pointer itself is an induction variable.
+ /// This check allows us to vectorize A[idx] into a wide load/store.
+ /// Returns:
+ /// 0 - Stride is unknown or non-consecutive.
+ /// 1 - Address is consecutive.
+ /// -1 - Address is consecutive, and decreasing.
+ int isConsecutivePtr(Value *Ptr);
- // Looking at the diagnostic output is the only way to determine if a loop
- // was vectorized (other than looking at the IR or machine code), so it
- // is important to generate an optimization remark for each loop. Most of
- // these messages are generated by emitOptimizationRemarkAnalysis. Remarks
- // generated by emitOptimizationRemark and emitOptimizationRemarkMissed are
- // less verbose reporting vectorized loops and unvectorized loops that may
- // benefit from vectorization, respectively.
+ /// Returns true if the value V is uniform within the loop.
+ bool isUniform(Value *V);
- if (Hints.getForce() == LoopVectorizeHints::FK_Disabled) {
- DEBUG(dbgs() << "LV: Not vectorizing: #pragma vectorize disable.\n");
- emitOptimizationRemarkAnalysis(F->getContext(), DEBUG_TYPE, *F,
- L->getStartLoc(), Hints.emitRemark());
- return false;
- }
+ /// Returns true if this instruction will remain scalar after vectorization.
+ bool isUniformAfterVectorization(Instruction* I) { return Uniforms.count(I); }
- if (!AlwaysVectorize && Hints.getForce() != LoopVectorizeHints::FK_Enabled) {
- DEBUG(dbgs() << "LV: Not vectorizing: No #pragma vectorize enable.\n");
- emitOptimizationRemarkAnalysis(F->getContext(), DEBUG_TYPE, *F,
- L->getStartLoc(), Hints.emitRemark());
- return false;
- }
+ /// Returns the information that we collected about runtime memory check.
+ const RuntimePointerChecking *getRuntimePointerChecking() const {
+ return LAI->getRuntimePointerChecking();
+ }
- if (Hints.getWidth() == 1 && Hints.getInterleave() == 1) {
- DEBUG(dbgs() << "LV: Not vectorizing: Disabled/already vectorized.\n");
- emitOptimizationRemarkAnalysis(
- F->getContext(), DEBUG_TYPE, *F, L->getStartLoc(),
- "loop not vectorized: vector width and interleave count are "
- "explicitly set to 1");
- return false;
- }
+ const LoopAccessInfo *getLAI() const {
+ return LAI;
+ }
- // Check the loop for a trip count threshold:
- // do not vectorize loops with a tiny trip count.
- const unsigned TC = SE->getSmallConstantTripCount(L);
- if (TC > 0u && TC < TinyTripCountVectorThreshold) {
- DEBUG(dbgs() << "LV: Found a loop with a very small trip count. "
- << "This loop is not worth vectorizing.");
- if (Hints.getForce() == LoopVectorizeHints::FK_Enabled)
- DEBUG(dbgs() << " But vectorizing was explicitly forced.\n");
- else {
- DEBUG(dbgs() << "\n");
- emitOptimizationRemarkAnalysis(
- F->getContext(), DEBUG_TYPE, *F, L->getStartLoc(),
- "vectorization is not beneficial and is not explicitly forced");
- return false;
- }
- }
+ /// \brief Check if \p Instr belongs to any interleaved access group.
+ bool isAccessInterleaved(Instruction *Instr) {
+ return InterleaveInfo.isInterleaved(Instr);
+ }
- // Check if it is legal to vectorize the loop.
- LoopVectorizationLegality LVL(L, SE, DL, DT, TLI, AA, F, TTI);
- if (!LVL.canVectorize()) {
- DEBUG(dbgs() << "LV: Not vectorizing: Cannot prove legality.\n");
- emitMissedWarning(F, L, Hints);
- return false;
- }
+ /// \brief Get the interleaved access group that \p Instr belongs to.
+ const InterleaveGroup *getInterleavedAccessGroup(Instruction *Instr) {
+ return InterleaveInfo.getInterleaveGroup(Instr);
+ }
- // Use the cost model.
- LoopVectorizationCostModel CM(L, SE, LI, &LVL, *TTI, DL, TLI, AC, F,
- &Hints);
+ unsigned getMaxSafeDepDistBytes() { return LAI->getMaxSafeDepDistBytes(); }
- // Check the function attributes to find out if this function should be
- // optimized for size.
- bool OptForSize = Hints.getForce() != LoopVectorizeHints::FK_Enabled &&
- F->hasFnAttribute(Attribute::OptimizeForSize);
+ bool hasStride(Value *V) { return StrideSet.count(V); }
+ bool mustCheckStrides() { return !StrideSet.empty(); }
+ SmallPtrSet<Value *, 8>::iterator strides_begin() {
+ return StrideSet.begin();
+ }
+ SmallPtrSet<Value *, 8>::iterator strides_end() { return StrideSet.end(); }
- // Compute the weighted frequency of this loop being executed and see if it
- // is less than 20% of the function entry baseline frequency. Note that we
- // always have a canonical loop here because we think we *can* vectoriez.
- // FIXME: This is hidden behind a flag due to pervasive problems with
- // exactly what block frequency models.
- if (LoopVectorizeWithBlockFrequency) {
- BlockFrequency LoopEntryFreq = BFI->getBlockFreq(L->getLoopPreheader());
- if (Hints.getForce() != LoopVectorizeHints::FK_Enabled &&
- LoopEntryFreq < ColdEntryFreq)
- OptForSize = true;
- }
+ /// Returns true if the target machine supports masked store operation
+ /// for the given \p DataType and kind of access to \p Ptr.
+ bool isLegalMaskedStore(Type *DataType, Value *Ptr) {
+ return TTI->isLegalMaskedStore(DataType, isConsecutivePtr(Ptr));
+ }
+ /// Returns true if the target machine supports masked load operation
+ /// for the given \p DataType and kind of access to \p Ptr.
+ bool isLegalMaskedLoad(Type *DataType, Value *Ptr) {
+ return TTI->isLegalMaskedLoad(DataType, isConsecutivePtr(Ptr));
+ }
+ /// Returns true if vector representation of the instruction \p I
+ /// requires mask.
+ bool isMaskRequired(const Instruction* I) {
+ return (MaskedOp.count(I) != 0);
+ }
+ unsigned getNumStores() const {
+ return LAI->getNumStores();
+ }
+ unsigned getNumLoads() const {
+ return LAI->getNumLoads();
+ }
+ unsigned getNumPredStores() const {
+ return NumPredStores;
+ }
+private:
+ /// Check if a single basic block loop is vectorizable.
+ /// At this point we know that this is a loop with a constant trip count
+ /// and we only need to check individual instructions.
+ bool canVectorizeInstrs();
- // Check the function attributes to see if implicit floats are allowed.a
- // FIXME: This check doesn't seem possibly correct -- what if the loop is
- // an integer loop and the vector instructions selected are purely integer
- // vector instructions?
- if (F->hasFnAttribute(Attribute::NoImplicitFloat)) {
- DEBUG(dbgs() << "LV: Can't vectorize when the NoImplicitFloat"
- "attribute is used.\n");
- emitOptimizationRemarkAnalysis(
- F->getContext(), DEBUG_TYPE, *F, L->getStartLoc(),
- "loop not vectorized due to NoImplicitFloat attribute");
- emitMissedWarning(F, L, Hints);
- return false;
- }
+ /// When we vectorize loops we may change the order in which
+ /// we read and write from memory. This method checks if it is
+ /// legal to vectorize the code, considering only memory constrains.
+ /// Returns true if the loop is vectorizable
+ bool canVectorizeMemory();
- // Select the optimal vectorization factor.
- const LoopVectorizationCostModel::VectorizationFactor VF =
- CM.selectVectorizationFactor(OptForSize);
+ /// Return true if we can vectorize this loop using the IF-conversion
+ /// transformation.
+ bool canVectorizeWithIfConvert();
- // Select the unroll factor.
- const unsigned UF =
- CM.selectUnrollFactor(OptForSize, VF.Width, VF.Cost);
+ /// Collect the variables that need to stay uniform after vectorization.
+ void collectLoopUniforms();
- DEBUG(dbgs() << "LV: Found a vectorizable loop (" << VF.Width << ") in "
- << DebugLocStr << '\n');
- DEBUG(dbgs() << "LV: Unroll Factor is " << UF << '\n');
+ /// Return true if all of the instructions in the block can be speculatively
+ /// executed. \p SafePtrs is a list of addresses that are known to be legal
+ /// and we know that we can read from them without segfault.
+ bool blockCanBePredicated(BasicBlock *BB, SmallPtrSetImpl<Value *> &SafePtrs);
- if (VF.Width == 1) {
- DEBUG(dbgs() << "LV: Vectorization is possible but not beneficial\n");
+ /// Returns the induction kind of Phi and record the step. This function may
+ /// return NoInduction if the PHI is not an induction variable.
+ InductionKind isInductionVariable(PHINode *Phi, ConstantInt *&StepValue);
- if (UF == 1) {
- emitOptimizationRemarkAnalysis(
- F->getContext(), DEBUG_TYPE, *F, L->getStartLoc(),
- "not beneficial to vectorize and user disabled interleaving");
- return false;
- }
- DEBUG(dbgs() << "LV: Trying to at least unroll the loops.\n");
+ /// \brief Collect memory access with loop invariant strides.
+ ///
+ /// Looks for accesses like "a[i * StrideA]" where "StrideA" is loop
+ /// invariant.
+ void collectStridedAccess(Value *LoadOrStoreInst);
- // Report the unrolling decision.
- emitOptimizationRemark(F->getContext(), DEBUG_TYPE, *F, L->getStartLoc(),
- Twine("unrolled with interleaving factor " +
- Twine(UF) +
- " (vectorization not beneficial)"));
+ /// Report an analysis message to assist the user in diagnosing loops that are
+ /// not vectorized. These are handled as LoopAccessReport rather than
+ /// VectorizationReport because the << operator of VectorizationReport returns
+ /// LoopAccessReport.
+ void emitAnalysis(const LoopAccessReport &Message) const {
+ emitAnalysisDiag(TheFunction, TheLoop, *Hints, Message);
+ }
- // We decided not to vectorize, but we may want to unroll.
+ unsigned NumPredStores;
- InnerLoopUnroller Unroller(L, SE, LI, DT, DL, TLI, UF);
- Unroller.vectorize(&LVL);
- } else {
- // If we decided that it is *legal* to vectorize the loop then do it.
- InnerLoopVectorizer LB(L, SE, LI, DT, DL, TLI, VF.Width, UF);
- LB.vectorize(&LVL);
- ++LoopsVectorized;
+ /// The loop that we evaluate.
+ Loop *TheLoop;
+ /// Scev analysis.
+ ScalarEvolution *SE;
+ /// Target Library Info.
+ TargetLibraryInfo *TLI;
+ /// Parent function
+ Function *TheFunction;
+ /// Target Transform Info
+ const TargetTransformInfo *TTI;
+ /// Dominator Tree.
+ DominatorTree *DT;
+ // LoopAccess analysis.
+ LoopAccessAnalysis *LAA;
+ // And the loop-accesses info corresponding to this loop. This pointer is
+ // null until canVectorizeMemory sets it up.
+ const LoopAccessInfo *LAI;
- // Report the vectorization decision.
- emitOptimizationRemark(
- F->getContext(), DEBUG_TYPE, *F, L->getStartLoc(),
- Twine("vectorized loop (vectorization factor: ") + Twine(VF.Width) +
- ", unrolling interleave factor: " + Twine(UF) + ")");
- }
+ /// The interleave access information contains groups of interleaved accesses
+ /// with the same stride and close to each other.
+ InterleavedAccessInfo InterleaveInfo;
- // Mark the loop as already vectorized to avoid vectorizing again.
- Hints.setAlreadyVectorized();
+ // --- vectorization state --- //
- DEBUG(verifyFunction(*L->getHeader()->getParent()));
- return true;
- }
+ /// Holds the integer induction variable. This is the counter of the
+ /// loop.
+ PHINode *Induction;
+ /// Holds the reduction variables.
+ ReductionList Reductions;
+ /// Holds all of the induction variables that we found in the loop.
+ /// Notice that inductions don't need to start at zero and that induction
+ /// variables can be pointers.
+ InductionList Inductions;
+ /// Holds the widest induction type encountered.
+ Type *WidestIndTy;
- void getAnalysisUsage(AnalysisUsage &AU) const override {
- AU.addRequired<AssumptionCacheTracker>();
- AU.addRequiredID(LoopSimplifyID);
- AU.addRequiredID(LCSSAID);
- AU.addRequired<BlockFrequencyInfo>();
- AU.addRequired<DominatorTreeWrapperPass>();
- AU.addRequired<LoopInfoWrapperPass>();
- AU.addRequired<ScalarEvolution>();
- AU.addRequired<TargetTransformInfoWrapperPass>();
- AU.addRequired<AliasAnalysis>();
- AU.addPreserved<LoopInfoWrapperPass>();
- AU.addPreserved<DominatorTreeWrapperPass>();
- AU.addPreserved<AliasAnalysis>();
- }
+ /// Allowed outside users. This holds the reduction
+ /// vars which can be accessed from outside the loop.
+ SmallPtrSet<Value*, 4> AllowedExit;
+ /// This set holds the variables which are known to be uniform after
+ /// vectorization.
+ SmallPtrSet<Instruction*, 4> Uniforms;
-};
+ /// Can we assume the absence of NaNs.
+ bool HasFunNoNaNAttr;
-} // end anonymous namespace
+ /// Vectorization requirements that will go through late-evaluation.
+ LoopVectorizationRequirements *Requirements;
-//===----------------------------------------------------------------------===//
-// Implementation of LoopVectorizationLegality, InnerLoopVectorizer and
-// LoopVectorizationCostModel.
-//===----------------------------------------------------------------------===//
+ /// Used to emit an analysis of any legality issues.
+ const LoopVectorizeHints *Hints;
-static Value *stripIntegerCast(Value *V) {
- if (CastInst *CI = dyn_cast<CastInst>(V))
- if (CI->getOperand(0)->getType()->isIntegerTy())
- return CI->getOperand(0);
- return V;
-}
+ ValueToValueMap Strides;
+ SmallPtrSet<Value *, 8> StrideSet;
-///\brief Replaces the symbolic stride in a pointer SCEV expression by one.
-///
-/// If \p OrigPtr is not null, use it to look up the stride value instead of
-/// \p Ptr.
-static const SCEV *replaceSymbolicStrideSCEV(ScalarEvolution *SE,
- ValueToValueMap &PtrToStride,
- Value *Ptr, Value *OrigPtr = nullptr) {
+ /// While vectorizing these instructions we have to generate a
+ /// call to the appropriate masked intrinsic
+ SmallPtrSet<const Instruction*, 8> MaskedOp;
+};
- const SCEV *OrigSCEV = SE->getSCEV(Ptr);
+/// LoopVectorizationCostModel - estimates the expected speedups due to
+/// vectorization.
+/// In many cases vectorization is not profitable. This can happen because of
+/// a number of reasons. In this class we mainly attempt to predict the
+/// expected speedup/slowdowns due to the supported instruction set. We use the
+/// TargetTransformInfo to query the different backends for the cost of
+/// different operations.
+class LoopVectorizationCostModel {
+public:
+ LoopVectorizationCostModel(Loop *L, ScalarEvolution *SE, LoopInfo *LI,
+ LoopVectorizationLegality *Legal,
+ const TargetTransformInfo &TTI,
+ const TargetLibraryInfo *TLI, AssumptionCache *AC,
+ const Function *F, const LoopVectorizeHints *Hints)
+ : TheLoop(L), SE(SE), LI(LI), Legal(Legal), TTI(TTI), TLI(TLI),
+ TheFunction(F), Hints(Hints) {
+ CodeMetrics::collectEphemeralValues(L, AC, EphValues);
+ }
- // If there is an entry in the map return the SCEV of the pointer with the
- // symbolic stride replaced by one.
- ValueToValueMap::iterator SI = PtrToStride.find(OrigPtr ? OrigPtr : Ptr);
- if (SI != PtrToStride.end()) {
- Value *StrideVal = SI->second;
+ /// Information about vectorization costs
+ struct VectorizationFactor {
+ unsigned Width; // Vector width with best cost
+ unsigned Cost; // Cost of the loop with that width
+ };
+ /// \return The most profitable vectorization factor and the cost of that VF.
+ /// This method checks every power of two up to VF. If UserVF is not ZERO
+ /// then this vectorization factor will be selected if vectorization is
+ /// possible.
+ VectorizationFactor selectVectorizationFactor(bool OptForSize);
- // Strip casts.
- StrideVal = stripIntegerCast(StrideVal);
+ /// \return The size (in bits) of the widest type in the code that
+ /// needs to be vectorized. We ignore values that remain scalar such as
+ /// 64 bit loop indices.
+ unsigned getWidestType();
- // Replace symbolic stride by one.
- Value *One = ConstantInt::get(StrideVal->getType(), 1);
- ValueToValueMap RewriteMap;
- RewriteMap[StrideVal] = One;
+ /// \return The desired interleave count.
+ /// If interleave count has been specified by metadata it will be returned.
+ /// Otherwise, the interleave count is computed and returned. VF and LoopCost
+ /// are the selected vectorization factor and the cost of the selected VF.
+ unsigned selectInterleaveCount(bool OptForSize, unsigned VF,
+ unsigned LoopCost);
- const SCEV *ByOne =
- SCEVParameterRewriter::rewrite(OrigSCEV, *SE, RewriteMap, true);
- DEBUG(dbgs() << "LV: Replacing SCEV: " << *OrigSCEV << " by: " << *ByOne
- << "\n");
- return ByOne;
- }
+ /// \return The most profitable unroll factor.
+ /// This method finds the best unroll-factor based on register pressure and
+ /// other parameters. VF and LoopCost are the selected vectorization factor
+ /// and the cost of the selected VF.
+ unsigned computeInterleaveCount(bool OptForSize, unsigned VF,
+ unsigned LoopCost);
- // Otherwise, just return the SCEV of the original pointer.
- return SE->getSCEV(Ptr);
-}
+ /// \brief A struct that represents some properties of the register usage
+ /// of a loop.
+ struct RegisterUsage {
+ /// Holds the number of loop invariant values that are used in the loop.
+ unsigned LoopInvariantRegs;
+ /// Holds the maximum number of concurrent live intervals in the loop.
+ unsigned MaxLocalUsers;
+ /// Holds the number of instructions in the loop.
+ unsigned NumInstructions;
+ };
-void RuntimePointerCheck::insert(ScalarEvolution *SE, Loop *Lp, Value *Ptr,
- bool WritePtr, unsigned DepSetId,
- unsigned ASId, ValueToValueMap &Strides) {
- // Get the stride replaced scev.
- const SCEV *Sc = replaceSymbolicStrideSCEV(SE, Strides, Ptr);
- const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(Sc);
- assert(AR && "Invalid addrec expression");
- const SCEV *Ex = SE->getBackedgeTakenCount(Lp);
- const SCEV *ScEnd = AR->evaluateAtIteration(Ex, *SE);
- Pointers.push_back(Ptr);
- Starts.push_back(AR->getStart());
- Ends.push_back(ScEnd);
- IsWritePtr.push_back(WritePtr);
- DependencySetId.push_back(DepSetId);
- AliasSetId.push_back(ASId);
-}
+ /// \return information about the register usage of the loop.
+ RegisterUsage calculateRegisterUsage();
-Value *InnerLoopVectorizer::getBroadcastInstrs(Value *V) {
- // We need to place the broadcast of invariant variables outside the loop.
- Instruction *Instr = dyn_cast<Instruction>(V);
- bool NewInstr =
- (Instr && std::find(LoopVectorBody.begin(), LoopVectorBody.end(),
- Instr->getParent()) != LoopVectorBody.end());
- bool Invariant = OrigLoop->isLoopInvariant(V) && !NewInstr;
+private:
+ /// Returns the expected execution cost. The unit of the cost does
+ /// not matter because we use the 'cost' units to compare different
+ /// vector widths. The cost that is returned is *not* normalized by
+ /// the factor width.
+ unsigned expectedCost(unsigned VF);
- // Place the code for broadcasting invariant variables in the new preheader.
- IRBuilder<>::InsertPointGuard Guard(Builder);
- if (Invariant)
- Builder.SetInsertPoint(LoopVectorPreHeader->getTerminator());
+ /// Returns the execution time cost of an instruction for a given vector
+ /// width. Vector width of one means scalar.
+ unsigned getInstructionCost(Instruction *I, unsigned VF);
- // Broadcast the scalar into all locations in the vector.
- Value *Shuf = Builder.CreateVectorSplat(VF, V, "broadcast");
+ /// Returns whether the instruction is a load or store and will be a emitted
+ /// as a vector operation.
+ bool isConsecutiveLoadOrStore(Instruction *I);
- return Shuf;
-}
+ /// Report an analysis message to assist the user in diagnosing loops that are
+ /// not vectorized. These are handled as LoopAccessReport rather than
+ /// VectorizationReport because the << operator of VectorizationReport returns
+ /// LoopAccessReport.
+ void emitAnalysis(const LoopAccessReport &Message) const {
+ emitAnalysisDiag(TheFunction, TheLoop, *Hints, Message);
+ }
-Value *InnerLoopVectorizer::getStepVector(Value *Val, int StartIdx,
- Value *Step) {
- assert(Val->getType()->isVectorTy() && "Must be a vector");
- assert(Val->getType()->getScalarType()->isIntegerTy() &&
- "Elem must be an integer");
- assert(Step->getType() == Val->getType()->getScalarType() &&
- "Step has wrong type");
- // Create the types.
- Type *ITy = Val->getType()->getScalarType();
- VectorType *Ty = cast<VectorType>(Val->getType());
- int VLen = Ty->getNumElements();
- SmallVector<Constant*, 8> Indices;
+ /// Values used only by @llvm.assume calls.
+ SmallPtrSet<const Value *, 32> EphValues;
- // Create a vector of consecutive numbers from zero to VF.
- for (int i = 0; i < VLen; ++i)
- Indices.push_back(ConstantInt::get(ITy, StartIdx + i));
+ /// The loop that we evaluate.
+ Loop *TheLoop;
+ /// Scev analysis.
+ ScalarEvolution *SE;
+ /// Loop Info analysis.
+ LoopInfo *LI;
+ /// Vectorization legality.
+ LoopVectorizationLegality *Legal;
+ /// Vector target information.
+ const TargetTransformInfo &TTI;
+ /// Target Library Info.
+ const TargetLibraryInfo *TLI;
+ const Function *TheFunction;
+ // Loop Vectorize Hint.
+ const LoopVectorizeHints *Hints;
+};
- // Add the consecutive indices to the vector value.
- Constant *Cv = ConstantVector::get(Indices);
- assert(Cv->getType() == Val->getType() && "Invalid consecutive vec");
- Step = Builder.CreateVectorSplat(VLen, Step);
- assert(Step->getType() == Val->getType() && "Invalid step vec");
- // FIXME: The newly created binary instructions should contain nsw/nuw flags,
- // which can be found from the original scalar operations.
- Step = Builder.CreateMul(Cv, Step);
- return Builder.CreateAdd(Val, Step, "induction");
-}
+/// \brief This holds vectorization requirements that must be verified late in
+/// the process. The requirements are set by legalize and costmodel. Once
+/// vectorization has been determined to be possible and profitable the
+/// requirements can be verified by looking for metadata or compiler options.
+/// For example, some loops require FP commutativity which is only allowed if
+/// vectorization is explicitly specified or if the fast-math compiler option
+/// has been provided.
+/// Late evaluation of these requirements allows helpful diagnostics to be
+/// composed that tells the user what need to be done to vectorize the loop. For
+/// example, by specifying #pragma clang loop vectorize or -ffast-math. Late
+/// evaluation should be used only when diagnostics can generated that can be
+/// followed by a non-expert user.
+class LoopVectorizationRequirements {
+public:
+ LoopVectorizationRequirements()
+ : NumRuntimePointerChecks(0), UnsafeAlgebraInst(nullptr) {}
+
+ void addUnsafeAlgebraInst(Instruction *I) {
+ // First unsafe algebra instruction.
+ if (!UnsafeAlgebraInst)
+ UnsafeAlgebraInst = I;
+ }
+
+ void addRuntimePointerChecks(unsigned Num) { NumRuntimePointerChecks = Num; }
+
+ bool doesNotMeet(Function *F, Loop *L, const LoopVectorizeHints &Hints) {
+ // If a loop hint is provided the diagnostic is always produced.
+ const char *Name = Hints.isForced() ? DiagnosticInfo::AlwaysPrint : LV_NAME;
+ bool failed = false;
+ if (UnsafeAlgebraInst &&
+ Hints.getForce() == LoopVectorizeHints::FK_Undefined &&
+ Hints.getWidth() == 0) {
+ emitOptimizationRemarkAnalysisFPCommute(
+ F->getContext(), Name, *F, UnsafeAlgebraInst->getDebugLoc(),
+ VectorizationReport() << "vectorization requires changes in the "
+ "order of operations, however IEEE 754 "
+ "floating-point operations are not "
+ "commutative");
+ failed = true;
+ }
-/// \brief Find the operand of the GEP that should be checked for consecutive
-/// stores. This ignores trailing indices that have no effect on the final
-/// pointer.
-static unsigned getGEPInductionOperand(const DataLayout *DL,
- const GetElementPtrInst *Gep) {
- unsigned LastOperand = Gep->getNumOperands() - 1;
- unsigned GEPAllocSize = DL->getTypeAllocSize(
- cast<PointerType>(Gep->getType()->getScalarType())->getElementType());
-
- // Walk backwards and try to peel off zeros.
- while (LastOperand > 1 && match(Gep->getOperand(LastOperand), m_Zero())) {
- // Find the type we're currently indexing into.
- gep_type_iterator GEPTI = gep_type_begin(Gep);
- std::advance(GEPTI, LastOperand - 1);
-
- // If it's a type with the same allocation size as the result of the GEP we
- // can peel off the zero index.
- if (DL->getTypeAllocSize(*GEPTI) != GEPAllocSize)
- break;
- --LastOperand;
+ if (NumRuntimePointerChecks >
+ VectorizerParams::RuntimeMemoryCheckThreshold) {
+ emitOptimizationRemarkAnalysisAliasing(
+ F->getContext(), Name, *F, L->getStartLoc(),
+ VectorizationReport()
+ << "cannot prove pointers refer to independent arrays in memory. "
+ "The loop requires "
+ << NumRuntimePointerChecks
+ << " runtime independence checks to vectorize the loop, but that "
+ "would exceed the limit of "
+ << VectorizerParams::RuntimeMemoryCheckThreshold << " checks");
+ DEBUG(dbgs() << "LV: Too many memory checks needed.\n");
+ failed = true;
+ }
+
+ return failed;
}
- return LastOperand;
+private:
+ unsigned NumRuntimePointerChecks;
+ Instruction *UnsafeAlgebraInst;
+};
+
+static void addInnerLoop(Loop &L, SmallVectorImpl<Loop *> &V) {
+ if (L.empty())
+ return V.push_back(&L);
+
+ for (Loop *InnerL : L)
+ addInnerLoop(*InnerL, V);
}
-int LoopVectorizationLegality::isConsecutivePtr(Value *Ptr) {
- assert(Ptr->getType()->isPointerTy() && "Unexpected non-ptr");
- // Make sure that the pointer does not point to structs.
- if (Ptr->getType()->getPointerElementType()->isAggregateType())
- return 0;
+/// The LoopVectorize Pass.
+struct LoopVectorize : public FunctionPass {
+ /// Pass identification, replacement for typeid
+ static char ID;
- // If this value is a pointer induction variable we know it is consecutive.
- PHINode *Phi = dyn_cast_or_null<PHINode>(Ptr);
- if (Phi && Inductions.count(Phi)) {
- InductionInfo II = Inductions[Phi];
- return II.getConsecutiveDirection();
+ explicit LoopVectorize(bool NoUnrolling = false, bool AlwaysVectorize = true)
+ : FunctionPass(ID),
+ DisableUnrolling(NoUnrolling),
+ AlwaysVectorize(AlwaysVectorize) {
+ initializeLoopVectorizePass(*PassRegistry::getPassRegistry());
}
- GetElementPtrInst *Gep = dyn_cast_or_null<GetElementPtrInst>(Ptr);
- if (!Gep)
- return 0;
+ ScalarEvolution *SE;
+ LoopInfo *LI;
+ TargetTransformInfo *TTI;
+ DominatorTree *DT;
+ BlockFrequencyInfo *BFI;
+ TargetLibraryInfo *TLI;
+ AliasAnalysis *AA;
+ AssumptionCache *AC;
+ LoopAccessAnalysis *LAA;
+ bool DisableUnrolling;
+ bool AlwaysVectorize;
- unsigned NumOperands = Gep->getNumOperands();
- Value *GpPtr = Gep->getPointerOperand();
- // If this GEP value is a consecutive pointer induction variable and all of
- // the indices are constant then we know it is consecutive. We can
- Phi = dyn_cast<PHINode>(GpPtr);
- if (Phi && Inductions.count(Phi)) {
+ BlockFrequency ColdEntryFreq;
- // Make sure that the pointer does not point to structs.
- PointerType *GepPtrType = cast<PointerType>(GpPtr->getType());
- if (GepPtrType->getElementType()->isAggregateType())
- return 0;
+ bool runOnFunction(Function &F) override {
+ SE = &getAnalysis<ScalarEvolutionWrapperPass>().getSE();
+ LI = &getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
+ TTI = &getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F);
+ DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
+ BFI = &getAnalysis<BlockFrequencyInfoWrapperPass>().getBFI();
+ auto *TLIP = getAnalysisIfAvailable<TargetLibraryInfoWrapperPass>();
+ TLI = TLIP ? &TLIP->getTLI() : nullptr;
+ AA = &getAnalysis<AliasAnalysis>();
+ AC = &getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
+ LAA = &getAnalysis<LoopAccessAnalysis>();
- // Make sure that all of the index operands are loop invariant.
- for (unsigned i = 1; i < NumOperands; ++i)
- if (!SE->isLoopInvariant(SE->getSCEV(Gep->getOperand(i)), TheLoop))
- return 0;
+ // Compute some weights outside of the loop over the loops. Compute this
+ // using a BranchProbability to re-use its scaling math.
+ const BranchProbability ColdProb(1, 5); // 20%
+ ColdEntryFreq = BlockFrequency(BFI->getEntryFreq()) * ColdProb;
- InductionInfo II = Inductions[Phi];
- return II.getConsecutiveDirection();
- }
+ // Don't attempt if
+ // 1. the target claims to have no vector registers, and
+ // 2. interleaving won't help ILP.
+ //
+ // The second condition is necessary because, even if the target has no
+ // vector registers, loop vectorization may still enable scalar
+ // interleaving.
+ if (!TTI->getNumberOfRegisters(true) && TTI->getMaxInterleaveFactor(1) < 2)
+ return false;
- unsigned InductionOperand = getGEPInductionOperand(DL, Gep);
+ // Build up a worklist of inner-loops to vectorize. This is necessary as
+ // the act of vectorizing or partially unrolling a loop creates new loops
+ // and can invalidate iterators across the loops.
+ SmallVector<Loop *, 8> Worklist;
- // Check that all of the gep indices are uniform except for our induction
- // operand.
- for (unsigned i = 0; i != NumOperands; ++i)
- if (i != InductionOperand &&
- !SE->isLoopInvariant(SE->getSCEV(Gep->getOperand(i)), TheLoop))
- return 0;
+ for (Loop *L : *LI)
+ addInnerLoop(*L, Worklist);
- // We can emit wide load/stores only if the last non-zero index is the
- // induction variable.
- const SCEV *Last = nullptr;
- if (!Strides.count(Gep))
- Last = SE->getSCEV(Gep->getOperand(InductionOperand));
- else {
- // Because of the multiplication by a stride we can have a s/zext cast.
- // We are going to replace this stride by 1 so the cast is safe to ignore.
- //
- // %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
- // %0 = trunc i64 %indvars.iv to i32
- // %mul = mul i32 %0, %Stride1
- // %idxprom = zext i32 %mul to i64 << Safe cast.
- // %arrayidx = getelementptr inbounds i32* %B, i64 %idxprom
- //
- Last = replaceSymbolicStrideSCEV(SE, Strides,
- Gep->getOperand(InductionOperand), Gep);
- if (const SCEVCastExpr *C = dyn_cast<SCEVCastExpr>(Last))
- Last =
- (C->getSCEVType() == scSignExtend || C->getSCEVType() == scZeroExtend)
- ? C->getOperand()
- : Last;
+ LoopsAnalyzed += Worklist.size();
+
+ // Now walk the identified inner loops.
+ bool Changed = false;
+ while (!Worklist.empty())
+ Changed |= processLoop(Worklist.pop_back_val());
+
+ // Process each loop nest in the function.
+ return Changed;
}
- if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(Last)) {
- const SCEV *Step = AR->getStepRecurrence(*SE);
- // The memory is consecutive because the last index is consecutive
- // and all other indices are loop invariant.
- if (Step->isOne())
- return 1;
- if (Step->isAllOnesValue())
- return -1;
+ static void AddRuntimeUnrollDisableMetaData(Loop *L) {
+ SmallVector<Metadata *, 4> MDs;
+ // Reserve first location for self reference to the LoopID metadata node.
+ MDs.push_back(nullptr);
+ bool IsUnrollMetadata = false;
+ MDNode *LoopID = L->getLoopID();
+ if (LoopID) {
+ // First find existing loop unrolling disable metadata.
+ for (unsigned i = 1, ie = LoopID->getNumOperands(); i < ie; ++i) {
+ MDNode *MD = dyn_cast<MDNode>(LoopID->getOperand(i));
+ if (MD) {
+ const MDString *S = dyn_cast<MDString>(MD->getOperand(0));
+ IsUnrollMetadata =
+ S && S->getString().startswith("llvm.loop.unroll.disable");
+ }
+ MDs.push_back(LoopID->getOperand(i));
+ }
+ }
+
+ if (!IsUnrollMetadata) {
+ // Add runtime unroll disable metadata.
+ LLVMContext &Context = L->getHeader()->getContext();
+ SmallVector<Metadata *, 1> DisableOperands;
+ DisableOperands.push_back(
+ MDString::get(Context, "llvm.loop.unroll.runtime.disable"));
+ MDNode *DisableNode = MDNode::get(Context, DisableOperands);
+ MDs.push_back(DisableNode);
+ MDNode *NewLoopID = MDNode::get(Context, MDs);
+ // Set operand 0 to refer to the loop id itself.
+ NewLoopID->replaceOperandWith(0, NewLoopID);
+ L->setLoopID(NewLoopID);
+ }
}
- return 0;
-}
+ bool processLoop(Loop *L) {
+ assert(L->empty() && "Only process inner loops.");
-bool LoopAccessAnalysis::isUniform(Value *V) {
- return (SE->isLoopInvariant(SE->getSCEV(V), TheLoop));
-}
+#ifndef NDEBUG
+ const std::string DebugLocStr = getDebugLocString(L);
+#endif /* NDEBUG */
-bool LoopVectorizationLegality::isUniform(Value *V) {
- return LAA.isUniform(V);
-}
+ DEBUG(dbgs() << "\nLV: Checking a loop in \""
+ << L->getHeader()->getParent()->getName() << "\" from "
+ << DebugLocStr << "\n");
-InnerLoopVectorizer::VectorParts&
-InnerLoopVectorizer::getVectorValue(Value *V) {
- assert(V != Induction && "The new induction variable should not be used.");
- assert(!V->getType()->isVectorTy() && "Can't widen a vector");
+ LoopVectorizeHints Hints(L, DisableUnrolling);
- // If we have a stride that is replaced by one, do it here.
- if (Legal->hasStride(V))
- V = ConstantInt::get(V->getType(), 1);
+ DEBUG(dbgs() << "LV: Loop hints:"
+ << " force="
+ << (Hints.getForce() == LoopVectorizeHints::FK_Disabled
+ ? "disabled"
+ : (Hints.getForce() == LoopVectorizeHints::FK_Enabled
+ ? "enabled"
+ : "?")) << " width=" << Hints.getWidth()
+ << " unroll=" << Hints.getInterleave() << "\n");
- // If we have this scalar in the map, return it.
- if (WidenMap.has(V))
- return WidenMap.get(V);
+ // Function containing loop
+ Function *F = L->getHeader()->getParent();
- // If this scalar is unknown, assume that it is a constant or that it is
- // loop invariant. Broadcast V and save the value for future uses.
- Value *B = getBroadcastInstrs(V);
- return WidenMap.splat(V, B);
-}
+ // Looking at the diagnostic output is the only way to determine if a loop
+ // was vectorized (other than looking at the IR or machine code), so it
+ // is important to generate an optimization remark for each loop. Most of
+ // these messages are generated by emitOptimizationRemarkAnalysis. Remarks
+ // generated by emitOptimizationRemark and emitOptimizationRemarkMissed are
+ // less verbose reporting vectorized loops and unvectorized loops that may
+ // benefit from vectorization, respectively.
-Value *InnerLoopVectorizer::reverseVector(Value *Vec) {
- assert(Vec->getType()->isVectorTy() && "Invalid type");
- SmallVector<Constant*, 8> ShuffleMask;
- for (unsigned i = 0; i < VF; ++i)
- ShuffleMask.push_back(Builder.getInt32(VF - i - 1));
+ if (!Hints.allowVectorization(F, L, AlwaysVectorize)) {
+ DEBUG(dbgs() << "LV: Loop hints prevent vectorization.\n");
+ return false;
+ }
- return Builder.CreateShuffleVector(Vec, UndefValue::get(Vec->getType()),
- ConstantVector::get(ShuffleMask),
- "reverse");
-}
+ // Check the loop for a trip count threshold:
+ // do not vectorize loops with a tiny trip count.
+ const unsigned TC = SE->getSmallConstantTripCount(L);
+ if (TC > 0u && TC < TinyTripCountVectorThreshold) {
+ DEBUG(dbgs() << "LV: Found a loop with a very small trip count. "
+ << "This loop is not worth vectorizing.");
+ if (Hints.getForce() == LoopVectorizeHints::FK_Enabled)
+ DEBUG(dbgs() << " But vectorizing was explicitly forced.\n");
+ else {
+ DEBUG(dbgs() << "\n");
+ emitAnalysisDiag(F, L, Hints, VectorizationReport()
+ << "vectorization is not beneficial "
+ "and is not explicitly forced");
+ return false;
+ }
+ }
-void InnerLoopVectorizer::vectorizeMemoryInstruction(Instruction *Instr) {
- // Attempt to issue a wide load.
- LoadInst *LI = dyn_cast<LoadInst>(Instr);
- StoreInst *SI = dyn_cast<StoreInst>(Instr);
+ // Check if it is legal to vectorize the loop.
+ LoopVectorizationRequirements Requirements;
+ LoopVectorizationLegality LVL(L, SE, DT, TLI, AA, F, TTI, LAA,
+ &Requirements, &Hints);
+ if (!LVL.canVectorize()) {
+ DEBUG(dbgs() << "LV: Not vectorizing: Cannot prove legality.\n");
+ emitMissedWarning(F, L, Hints);
+ return false;
+ }
- assert((LI || SI) && "Invalid Load/Store instruction");
+ // Use the cost model.
+ LoopVectorizationCostModel CM(L, SE, LI, &LVL, *TTI, TLI, AC, F, &Hints);
- Type *ScalarDataTy = LI ? LI->getType() : SI->getValueOperand()->getType();
- Type *DataTy = VectorType::get(ScalarDataTy, VF);
- Value *Ptr = LI ? LI->getPointerOperand() : SI->getPointerOperand();
- unsigned Alignment = LI ? LI->getAlignment() : SI->getAlignment();
- // An alignment of 0 means target abi alignment. We need to use the scalar's
- // target abi alignment in such a case.
- if (!Alignment)
- Alignment = DL->getABITypeAlignment(ScalarDataTy);
- unsigned AddressSpace = Ptr->getType()->getPointerAddressSpace();
- unsigned ScalarAllocatedSize = DL->getTypeAllocSize(ScalarDataTy);
- unsigned VectorElementSize = DL->getTypeStoreSize(DataTy)/VF;
+ // Check the function attributes to find out if this function should be
+ // optimized for size.
+ bool OptForSize = Hints.getForce() != LoopVectorizeHints::FK_Enabled &&
+ F->optForSize();
- if (SI && Legal->blockNeedsPredication(SI->getParent()) &&
- !Legal->isMaskRequired(SI))
- return scalarizeInstruction(Instr, true);
+ // Compute the weighted frequency of this loop being executed and see if it
+ // is less than 20% of the function entry baseline frequency. Note that we
+ // always have a canonical loop here because we think we *can* vectorize.
+ // FIXME: This is hidden behind a flag due to pervasive problems with
+ // exactly what block frequency models.
+ if (LoopVectorizeWithBlockFrequency) {
+ BlockFrequency LoopEntryFreq = BFI->getBlockFreq(L->getLoopPreheader());
+ if (Hints.getForce() != LoopVectorizeHints::FK_Enabled &&
+ LoopEntryFreq < ColdEntryFreq)
+ OptForSize = true;
+ }
- if (ScalarAllocatedSize != VectorElementSize)
- return scalarizeInstruction(Instr);
+ // Check the function attributes to see if implicit floats are allowed.
+ // FIXME: This check doesn't seem possibly correct -- what if the loop is
+ // an integer loop and the vector instructions selected are purely integer
+ // vector instructions?
+ if (F->hasFnAttribute(Attribute::NoImplicitFloat)) {
+ DEBUG(dbgs() << "LV: Can't vectorize when the NoImplicitFloat"
+ "attribute is used.\n");
+ emitAnalysisDiag(
+ F, L, Hints,
+ VectorizationReport()
+ << "loop not vectorized due to NoImplicitFloat attribute");
+ emitMissedWarning(F, L, Hints);
+ return false;
+ }
- // If the pointer is loop invariant or if it is non-consecutive,
- // scalarize the load.
- int ConsecutiveStride = Legal->isConsecutivePtr(Ptr);
- bool Reverse = ConsecutiveStride < 0;
- bool UniformLoad = LI && Legal->isUniform(Ptr);
- if (!ConsecutiveStride || UniformLoad)
- return scalarizeInstruction(Instr);
+ // Select the optimal vectorization factor.
+ const LoopVectorizationCostModel::VectorizationFactor VF =
+ CM.selectVectorizationFactor(OptForSize);
- Constant *Zero = Builder.getInt32(0);
- VectorParts &Entry = WidenMap.get(Instr);
+ // Select the interleave count.
+ unsigned IC = CM.selectInterleaveCount(OptForSize, VF.Width, VF.Cost);
- // Handle consecutive loads/stores.
- GetElementPtrInst *Gep = dyn_cast<GetElementPtrInst>(Ptr);
- if (Gep && Legal->isInductionVariable(Gep->getPointerOperand())) {
- setDebugLocFromInst(Builder, Gep);
- Value *PtrOperand = Gep->getPointerOperand();
- Value *FirstBasePtr = getVectorValue(PtrOperand)[0];
- FirstBasePtr = Builder.CreateExtractElement(FirstBasePtr, Zero);
+ // Get user interleave count.
+ unsigned UserIC = Hints.getInterleave();
- // Create the new GEP with the new induction variable.
- GetElementPtrInst *Gep2 = cast<GetElementPtrInst>(Gep->clone());
- Gep2->setOperand(0, FirstBasePtr);
- Gep2->setName("gep.indvar.base");
- Ptr = Builder.Insert(Gep2);
- } else if (Gep) {
- setDebugLocFromInst(Builder, Gep);
- assert(SE->isLoopInvariant(SE->getSCEV(Gep->getPointerOperand()),
- OrigLoop) && "Base ptr must be invariant");
+ // Identify the diagnostic messages that should be produced.
+ std::string VecDiagMsg, IntDiagMsg;
+ bool VectorizeLoop = true, InterleaveLoop = true;
- // The last index does not have to be the induction. It can be
- // consecutive and be a function of the index. For example A[I+1];
- unsigned NumOperands = Gep->getNumOperands();
- unsigned InductionOperand = getGEPInductionOperand(DL, Gep);
- // Create the new GEP with the new induction variable.
- GetElementPtrInst *Gep2 = cast<GetElementPtrInst>(Gep->clone());
+ if (Requirements.doesNotMeet(F, L, Hints)) {
+ DEBUG(dbgs() << "LV: Not vectorizing: loop did not meet vectorization "
+ "requirements.\n");
+ emitMissedWarning(F, L, Hints);
+ return false;
+ }
- for (unsigned i = 0; i < NumOperands; ++i) {
- Value *GepOperand = Gep->getOperand(i);
- Instruction *GepOperandInst = dyn_cast<Instruction>(GepOperand);
+ if (VF.Width == 1) {
+ DEBUG(dbgs() << "LV: Vectorization is possible but not beneficial.\n");
+ VecDiagMsg =
+ "the cost-model indicates that vectorization is not beneficial";
+ VectorizeLoop = false;
+ }
- // Update last index or loop invariant instruction anchored in loop.
- if (i == InductionOperand ||
- (GepOperandInst && OrigLoop->contains(GepOperandInst))) {
- assert((i == InductionOperand ||
- SE->isLoopInvariant(SE->getSCEV(GepOperandInst), OrigLoop)) &&
- "Must be last index or loop invariant");
+ if (IC == 1 && UserIC <= 1) {
+ // Tell the user interleaving is not beneficial.
+ DEBUG(dbgs() << "LV: Interleaving is not beneficial.\n");
+ IntDiagMsg =
+ "the cost-model indicates that interleaving is not beneficial";
+ InterleaveLoop = false;
+ if (UserIC == 1)
+ IntDiagMsg +=
+ " and is explicitly disabled or interleave count is set to 1";
+ } else if (IC > 1 && UserIC == 1) {
+ // Tell the user interleaving is beneficial, but it explicitly disabled.
+ DEBUG(dbgs()
+ << "LV: Interleaving is beneficial but is explicitly disabled.");
+ IntDiagMsg = "the cost-model indicates that interleaving is beneficial "
+ "but is explicitly disabled or interleave count is set to 1";
+ InterleaveLoop = false;
+ }
- VectorParts &GEPParts = getVectorValue(GepOperand);
- Value *Index = GEPParts[0];
- Index = Builder.CreateExtractElement(Index, Zero);
- Gep2->setOperand(i, Index);
- Gep2->setName("gep.indvar.idx");
- }
+ // Override IC if user provided an interleave count.
+ IC = UserIC > 0 ? UserIC : IC;
+
+ // Emit diagnostic messages, if any.
+ if (!VectorizeLoop && !InterleaveLoop) {
+ // Do not vectorize or interleaving the loop.
+ emitOptimizationRemarkAnalysis(F->getContext(), DEBUG_TYPE, *F,
+ L->getStartLoc(), VecDiagMsg);
+ emitOptimizationRemarkAnalysis(F->getContext(), DEBUG_TYPE, *F,
+ L->getStartLoc(), IntDiagMsg);
+ return false;
+ } else if (!VectorizeLoop && InterleaveLoop) {
+ DEBUG(dbgs() << "LV: Interleave Count is " << IC << '\n');
+ emitOptimizationRemarkAnalysis(F->getContext(), DEBUG_TYPE, *F,
+ L->getStartLoc(), VecDiagMsg);
+ } else if (VectorizeLoop && !InterleaveLoop) {
+ DEBUG(dbgs() << "LV: Found a vectorizable loop (" << VF.Width << ") in "
+ << DebugLocStr << '\n');
+ emitOptimizationRemarkAnalysis(F->getContext(), DEBUG_TYPE, *F,
+ L->getStartLoc(), IntDiagMsg);
+ } else if (VectorizeLoop && InterleaveLoop) {
+ DEBUG(dbgs() << "LV: Found a vectorizable loop (" << VF.Width << ") in "
+ << DebugLocStr << '\n');
+ DEBUG(dbgs() << "LV: Interleave Count is " << IC << '\n');
}
- Ptr = Builder.Insert(Gep2);
- } else {
- // Use the induction element ptr.
- assert(isa<PHINode>(Ptr) && "Invalid induction ptr");
- setDebugLocFromInst(Builder, Ptr);
- VectorParts &PtrVal = getVectorValue(Ptr);
- Ptr = Builder.CreateExtractElement(PtrVal[0], Zero);
- }
- VectorParts Mask = createBlockInMask(Instr->getParent());
- // Handle Stores:
- if (SI) {
- assert(!Legal->isUniform(SI->getPointerOperand()) &&
- "We do not allow storing to uniform addresses");
- setDebugLocFromInst(Builder, SI);
- // We don't want to update the value in the map as it might be used in
- // another expression. So don't use a reference type for "StoredVal".
- VectorParts StoredVal = getVectorValue(SI->getValueOperand());
-
- for (unsigned Part = 0; Part < UF; ++Part) {
- // Calculate the pointer for the specific unroll-part.
- Value *PartPtr = Builder.CreateGEP(Ptr, Builder.getInt32(Part * VF));
+ if (!VectorizeLoop) {
+ assert(IC > 1 && "interleave count should not be 1 or 0");
+ // If we decided that it is not legal to vectorize the loop then
+ // interleave it.
+ InnerLoopUnroller Unroller(L, SE, LI, DT, TLI, TTI, IC);
+ Unroller.vectorize(&LVL);
- if (Reverse) {
- // If we store to reverse consecutive memory locations then we need
- // to reverse the order of elements in the stored value.
- StoredVal[Part] = reverseVector(StoredVal[Part]);
- // If the address is consecutive but reversed, then the
- // wide store needs to start at the last vector element.
- PartPtr = Builder.CreateGEP(Ptr, Builder.getInt32(-Part * VF));
- PartPtr = Builder.CreateGEP(PartPtr, Builder.getInt32(1 - VF));
- Mask[Part] = reverseVector(Mask[Part]);
- }
+ emitOptimizationRemark(F->getContext(), DEBUG_TYPE, *F, L->getStartLoc(),
+ Twine("interleaved loop (interleaved count: ") +
+ Twine(IC) + ")");
+ } else {
+ // If we decided that it is *legal* to vectorize the loop then do it.
+ InnerLoopVectorizer LB(L, SE, LI, DT, TLI, TTI, VF.Width, IC);
+ LB.vectorize(&LVL);
+ ++LoopsVectorized;
- Value *VecPtr = Builder.CreateBitCast(PartPtr,
- DataTy->getPointerTo(AddressSpace));
+ // Add metadata to disable runtime unrolling scalar loop when there's no
+ // runtime check about strides and memory. Because at this situation,
+ // scalar loop is rarely used not worthy to be unrolled.
+ if (!LB.IsSafetyChecksAdded())
+ AddRuntimeUnrollDisableMetaData(L);
- Instruction *NewSI;
- if (Legal->isMaskRequired(SI))
- NewSI = Builder.CreateMaskedStore(StoredVal[Part], VecPtr, Alignment,
- Mask[Part]);
- else
- NewSI = Builder.CreateAlignedStore(StoredVal[Part], VecPtr, Alignment);
- propagateMetadata(NewSI, SI);
+ // Report the vectorization decision.
+ emitOptimizationRemark(F->getContext(), DEBUG_TYPE, *F, L->getStartLoc(),
+ Twine("vectorized loop (vectorization width: ") +
+ Twine(VF.Width) + ", interleaved count: " +
+ Twine(IC) + ")");
}
- return;
- }
- // Handle loads.
- assert(LI && "Must have a load instruction");
- setDebugLocFromInst(Builder, LI);
- for (unsigned Part = 0; Part < UF; ++Part) {
- // Calculate the pointer for the specific unroll-part.
- Value *PartPtr = Builder.CreateGEP(Ptr, Builder.getInt32(Part * VF));
+ // Mark the loop as already vectorized to avoid vectorizing again.
+ Hints.setAlreadyVectorized();
- if (Reverse) {
- // If the address is consecutive but reversed, then the
- // wide load needs to start at the last vector element.
- PartPtr = Builder.CreateGEP(Ptr, Builder.getInt32(-Part * VF));
- PartPtr = Builder.CreateGEP(PartPtr, Builder.getInt32(1 - VF));
- Mask[Part] = reverseVector(Mask[Part]);
- }
+ DEBUG(verifyFunction(*L->getHeader()->getParent()));
+ return true;
+ }
- Instruction* NewLI;
- Value *VecPtr = Builder.CreateBitCast(PartPtr,
- DataTy->getPointerTo(AddressSpace));
- if (Legal->isMaskRequired(LI))
- NewLI = Builder.CreateMaskedLoad(VecPtr, Alignment, Mask[Part],
- UndefValue::get(DataTy),
- "wide.masked.load");
- else
- NewLI = Builder.CreateAlignedLoad(VecPtr, Alignment, "wide.load");
- propagateMetadata(NewLI, LI);
- Entry[Part] = Reverse ? reverseVector(NewLI) : NewLI;
+ void getAnalysisUsage(AnalysisUsage &AU) const override {
+ AU.addRequired<AssumptionCacheTracker>();
+ AU.addRequiredID(LoopSimplifyID);
+ AU.addRequiredID(LCSSAID);
+ AU.addRequired<BlockFrequencyInfoWrapperPass>();
+ AU.addRequired<DominatorTreeWrapperPass>();
+ AU.addRequired<LoopInfoWrapperPass>();
+ AU.addRequired<ScalarEvolutionWrapperPass>();
+ AU.addRequired<TargetTransformInfoWrapperPass>();
+ AU.addRequired<AliasAnalysis>();
+ AU.addRequired<LoopAccessAnalysis>();
+ AU.addPreserved<LoopInfoWrapperPass>();
+ AU.addPreserved<DominatorTreeWrapperPass>();
+ AU.addPreserved<AliasAnalysis>();
}
-}
-void InnerLoopVectorizer::scalarizeInstruction(Instruction *Instr, bool IfPredicateStore) {
- assert(!Instr->getType()->isAggregateType() && "Can't handle vectors");
- // Holds vector parameters or scalars, in case of uniform vals.
- SmallVector<VectorParts, 4> Params;
+};
- setDebugLocFromInst(Builder, Instr);
+} // end anonymous namespace
- // Find all of the vectorized parameters.
- for (unsigned op = 0, e = Instr->getNumOperands(); op != e; ++op) {
- Value *SrcOp = Instr->getOperand(op);
+//===----------------------------------------------------------------------===//
+// Implementation of LoopVectorizationLegality, InnerLoopVectorizer and
+// LoopVectorizationCostModel.
+//===----------------------------------------------------------------------===//
- // If we are accessing the old induction variable, use the new one.
- if (SrcOp == OldInduction) {
- Params.push_back(getVectorValue(SrcOp));
- continue;
- }
+Value *InnerLoopVectorizer::getBroadcastInstrs(Value *V) {
+ // We need to place the broadcast of invariant variables outside the loop.
+ Instruction *Instr = dyn_cast<Instruction>(V);
+ bool NewInstr =
+ (Instr && std::find(LoopVectorBody.begin(), LoopVectorBody.end(),
+ Instr->getParent()) != LoopVectorBody.end());
+ bool Invariant = OrigLoop->isLoopInvariant(V) && !NewInstr;
- // Try using previously calculated values.
- Instruction *SrcInst = dyn_cast<Instruction>(SrcOp);
+ // Place the code for broadcasting invariant variables in the new preheader.
+ IRBuilder<>::InsertPointGuard Guard(Builder);
+ if (Invariant)
+ Builder.SetInsertPoint(LoopVectorPreHeader->getTerminator());
- // If the src is an instruction that appeared earlier in the basic block
- // then it should already be vectorized.
- if (SrcInst && OrigLoop->contains(SrcInst)) {
- assert(WidenMap.has(SrcInst) && "Source operand is unavailable");
- // The parameter is a vector value from earlier.
- Params.push_back(WidenMap.get(SrcInst));
- } else {
- // The parameter is a scalar from outside the loop. Maybe even a constant.
- VectorParts Scalars;
- Scalars.append(UF, SrcOp);
- Params.push_back(Scalars);
- }
- }
+ // Broadcast the scalar into all locations in the vector.
+ Value *Shuf = Builder.CreateVectorSplat(VF, V, "broadcast");
- assert(Params.size() == Instr->getNumOperands() &&
- "Invalid number of operands");
+ return Shuf;
+}
- // Does this instruction return a value ?
- bool IsVoidRetTy = Instr->getType()->isVoidTy();
+Value *InnerLoopVectorizer::getStepVector(Value *Val, int StartIdx,
+ Value *Step) {
+ assert(Val->getType()->isVectorTy() && "Must be a vector");
+ assert(Val->getType()->getScalarType()->isIntegerTy() &&
+ "Elem must be an integer");
+ assert(Step->getType() == Val->getType()->getScalarType() &&
+ "Step has wrong type");
+ // Create the types.
+ Type *ITy = Val->getType()->getScalarType();
+ VectorType *Ty = cast<VectorType>(Val->getType());
+ int VLen = Ty->getNumElements();
+ SmallVector<Constant*, 8> Indices;
- Value *UndefVec = IsVoidRetTy ? nullptr :
- UndefValue::get(VectorType::get(Instr->getType(), VF));
- // Create a new entry in the WidenMap and initialize it to Undef or Null.
- VectorParts &VecResults = WidenMap.splat(Instr, UndefVec);
+ // Create a vector of consecutive numbers from zero to VF.
+ for (int i = 0; i < VLen; ++i)
+ Indices.push_back(ConstantInt::get(ITy, StartIdx + i));
- Instruction *InsertPt = Builder.GetInsertPoint();
- BasicBlock *IfBlock = Builder.GetInsertBlock();
- BasicBlock *CondBlock = nullptr;
+ // Add the consecutive indices to the vector value.
+ Constant *Cv = ConstantVector::get(Indices);
+ assert(Cv->getType() == Val->getType() && "Invalid consecutive vec");
+ Step = Builder.CreateVectorSplat(VLen, Step);
+ assert(Step->getType() == Val->getType() && "Invalid step vec");
+ // FIXME: The newly created binary instructions should contain nsw/nuw flags,
+ // which can be found from the original scalar operations.
+ Step = Builder.CreateMul(Cv, Step);
+ return Builder.CreateAdd(Val, Step, "induction");
+}
- VectorParts Cond;
- Loop *VectorLp = nullptr;
- if (IfPredicateStore) {
- assert(Instr->getParent()->getSinglePredecessor() &&
- "Only support single predecessor blocks");
- Cond = createEdgeMask(Instr->getParent()->getSinglePredecessor(),
- Instr->getParent());
- VectorLp = LI->getLoopFor(IfBlock);
- assert(VectorLp && "Must have a loop for this block");
+int LoopVectorizationLegality::isConsecutivePtr(Value *Ptr) {
+ assert(Ptr->getType()->isPointerTy() && "Unexpected non-ptr");
+ // Make sure that the pointer does not point to structs.
+ if (Ptr->getType()->getPointerElementType()->isAggregateType())
+ return 0;
+
+ // If this value is a pointer induction variable we know it is consecutive.
+ PHINode *Phi = dyn_cast_or_null<PHINode>(Ptr);
+ if (Phi && Inductions.count(Phi)) {
+ InductionInfo II = Inductions[Phi];
+ return II.getConsecutiveDirection();
}
- // For each vector unroll 'part':
- for (unsigned Part = 0; Part < UF; ++Part) {
- // For each scalar that we create:
- for (unsigned Width = 0; Width < VF; ++Width) {
+ GetElementPtrInst *Gep = dyn_cast_or_null<GetElementPtrInst>(Ptr);
+ if (!Gep)
+ return 0;
- // Start if-block.
- Value *Cmp = nullptr;
- if (IfPredicateStore) {
- Cmp = Builder.CreateExtractElement(Cond[Part], Builder.getInt32(Width));
- Cmp = Builder.CreateICmp(ICmpInst::ICMP_EQ, Cmp, ConstantInt::get(Cmp->getType(), 1));
- CondBlock = IfBlock->splitBasicBlock(InsertPt, "cond.store");
- LoopVectorBody.push_back(CondBlock);
- VectorLp->addBasicBlockToLoop(CondBlock, *LI);
- // Update Builder with newly created basic block.
- Builder.SetInsertPoint(InsertPt);
- }
+ unsigned NumOperands = Gep->getNumOperands();
+ Value *GpPtr = Gep->getPointerOperand();
+ // If this GEP value is a consecutive pointer induction variable and all of
+ // the indices are constant then we know it is consecutive. We can
+ Phi = dyn_cast<PHINode>(GpPtr);
+ if (Phi && Inductions.count(Phi)) {
- Instruction *Cloned = Instr->clone();
- if (!IsVoidRetTy)
- Cloned->setName(Instr->getName() + ".cloned");
- // Replace the operands of the cloned instructions with extracted scalars.
- for (unsigned op = 0, e = Instr->getNumOperands(); op != e; ++op) {
- Value *Op = Params[op][Part];
- // Param is a vector. Need to extract the right lane.
- if (Op->getType()->isVectorTy())
- Op = Builder.CreateExtractElement(Op, Builder.getInt32(Width));
- Cloned->setOperand(op, Op);
- }
+ // Make sure that the pointer does not point to structs.
+ PointerType *GepPtrType = cast<PointerType>(GpPtr->getType());
+ if (GepPtrType->getElementType()->isAggregateType())
+ return 0;
- // Place the cloned scalar in the new loop.
- Builder.Insert(Cloned);
+ // Make sure that all of the index operands are loop invariant.
+ for (unsigned i = 1; i < NumOperands; ++i)
+ if (!SE->isLoopInvariant(SE->getSCEV(Gep->getOperand(i)), TheLoop))
+ return 0;
- // If the original scalar returns a value we need to place it in a vector
- // so that future users will be able to use it.
- if (!IsVoidRetTy)
- VecResults[Part] = Builder.CreateInsertElement(VecResults[Part], Cloned,
- Builder.getInt32(Width));
- // End if-block.
- if (IfPredicateStore) {
- BasicBlock *NewIfBlock = CondBlock->splitBasicBlock(InsertPt, "else");
- LoopVectorBody.push_back(NewIfBlock);
- VectorLp->addBasicBlockToLoop(NewIfBlock, *LI);
- Builder.SetInsertPoint(InsertPt);
- Instruction *OldBr = IfBlock->getTerminator();
- BranchInst::Create(CondBlock, NewIfBlock, Cmp, OldBr);
- OldBr->eraseFromParent();
- IfBlock = NewIfBlock;
- }
- }
+ InductionInfo II = Inductions[Phi];
+ return II.getConsecutiveDirection();
}
-}
-static Instruction *getFirstInst(Instruction *FirstInst, Value *V,
- Instruction *Loc) {
- if (FirstInst)
- return FirstInst;
- if (Instruction *I = dyn_cast<Instruction>(V))
- return I->getParent() == Loc->getParent() ? I : nullptr;
- return nullptr;
-}
+ unsigned InductionOperand = getGEPInductionOperand(Gep);
-std::pair<Instruction *, Instruction *>
-InnerLoopVectorizer::addStrideCheck(Instruction *Loc) {
- Instruction *tnullptr = nullptr;
- if (!Legal->mustCheckStrides())
- return std::pair<Instruction *, Instruction *>(tnullptr, tnullptr);
+ // Check that all of the gep indices are uniform except for our induction
+ // operand.
+ for (unsigned i = 0; i != NumOperands; ++i)
+ if (i != InductionOperand &&
+ !SE->isLoopInvariant(SE->getSCEV(Gep->getOperand(i)), TheLoop))
+ return 0;
- IRBuilder<> ChkBuilder(Loc);
+ // We can emit wide load/stores only if the last non-zero index is the
+ // induction variable.
+ const SCEV *Last = nullptr;
+ if (!Strides.count(Gep))
+ Last = SE->getSCEV(Gep->getOperand(InductionOperand));
+ else {
+ // Because of the multiplication by a stride we can have a s/zext cast.
+ // We are going to replace this stride by 1 so the cast is safe to ignore.
+ //
+ // %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
+ // %0 = trunc i64 %indvars.iv to i32
+ // %mul = mul i32 %0, %Stride1
+ // %idxprom = zext i32 %mul to i64 << Safe cast.
+ // %arrayidx = getelementptr inbounds i32* %B, i64 %idxprom
+ //
+ Last = replaceSymbolicStrideSCEV(SE, Strides,
+ Gep->getOperand(InductionOperand), Gep);
+ if (const SCEVCastExpr *C = dyn_cast<SCEVCastExpr>(Last))
+ Last =
+ (C->getSCEVType() == scSignExtend || C->getSCEVType() == scZeroExtend)
+ ? C->getOperand()
+ : Last;
+ }
+ if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(Last)) {
+ const SCEV *Step = AR->getStepRecurrence(*SE);
- // Emit checks.
- Value *Check = nullptr;
- Instruction *FirstInst = nullptr;
- for (SmallPtrSet<Value *, 8>::iterator SI = Legal->strides_begin(),
- SE = Legal->strides_end();
- SI != SE; ++SI) {
- Value *Ptr = stripIntegerCast(*SI);
- Value *C = ChkBuilder.CreateICmpNE(Ptr, ConstantInt::get(Ptr->getType(), 1),
- "stride.chk");
- // Store the first instruction we create.
- FirstInst = getFirstInst(FirstInst, C, Loc);
- if (Check)
- Check = ChkBuilder.CreateOr(Check, C);
- else
- Check = C;
+ // The memory is consecutive because the last index is consecutive
+ // and all other indices are loop invariant.
+ if (Step->isOne())
+ return 1;
+ if (Step->isAllOnesValue())
+ return -1;
}
- // We have to do this trickery because the IRBuilder might fold the check to a
- // constant expression in which case there is no Instruction anchored in a
- // the block.
- LLVMContext &Ctx = Loc->getContext();
- Instruction *TheCheck =
- BinaryOperator::CreateAnd(Check, ConstantInt::getTrue(Ctx));
- ChkBuilder.Insert(TheCheck, "stride.not.one");
- FirstInst = getFirstInst(FirstInst, TheCheck, Loc);
+ return 0;
+}
- return std::make_pair(FirstInst, TheCheck);
+bool LoopVectorizationLegality::isUniform(Value *V) {
+ return LAI->isUniform(V);
}
-std::pair<Instruction *, Instruction *>
-InnerLoopVectorizer::addRuntimeCheck(Instruction *Loc) {
- RuntimePointerCheck *PtrRtCheck = Legal->getRuntimePointerCheck();
+InnerLoopVectorizer::VectorParts&
+InnerLoopVectorizer::getVectorValue(Value *V) {
+ assert(V != Induction && "The new induction variable should not be used.");
+ assert(!V->getType()->isVectorTy() && "Can't widen a vector");
- Instruction *tnullptr = nullptr;
- if (!PtrRtCheck->Need)
- return std::pair<Instruction *, Instruction *>(tnullptr, tnullptr);
+ // If we have a stride that is replaced by one, do it here.
+ if (Legal->hasStride(V))
+ V = ConstantInt::get(V->getType(), 1);
- unsigned NumPointers = PtrRtCheck->Pointers.size();
- SmallVector<TrackingVH<Value> , 2> Starts;
- SmallVector<TrackingVH<Value> , 2> Ends;
+ // If we have this scalar in the map, return it.
+ if (WidenMap.has(V))
+ return WidenMap.get(V);
- LLVMContext &Ctx = Loc->getContext();
- SCEVExpander Exp(*SE, "induction");
- Instruction *FirstInst = nullptr;
+ // If this scalar is unknown, assume that it is a constant or that it is
+ // loop invariant. Broadcast V and save the value for future uses.
+ Value *B = getBroadcastInstrs(V);
+ return WidenMap.splat(V, B);
+}
- for (unsigned i = 0; i < NumPointers; ++i) {
- Value *Ptr = PtrRtCheck->Pointers[i];
- const SCEV *Sc = SE->getSCEV(Ptr);
+Value *InnerLoopVectorizer::reverseVector(Value *Vec) {
+ assert(Vec->getType()->isVectorTy() && "Invalid type");
+ SmallVector<Constant*, 8> ShuffleMask;
+ for (unsigned i = 0; i < VF; ++i)
+ ShuffleMask.push_back(Builder.getInt32(VF - i - 1));
- if (SE->isLoopInvariant(Sc, OrigLoop)) {
- DEBUG(dbgs() << "LV: Adding RT check for a loop invariant ptr:" <<
- *Ptr <<"\n");
- Starts.push_back(Ptr);
- Ends.push_back(Ptr);
- } else {
- DEBUG(dbgs() << "LV: Adding RT check for range:" << *Ptr << '\n');
- unsigned AS = Ptr->getType()->getPointerAddressSpace();
+ return Builder.CreateShuffleVector(Vec, UndefValue::get(Vec->getType()),
+ ConstantVector::get(ShuffleMask),
+ "reverse");
+}
- // Use this type for pointer arithmetic.
- Type *PtrArithTy = Type::getInt8PtrTy(Ctx, AS);
+// Get a mask to interleave \p NumVec vectors into a wide vector.
+// I.e. <0, VF, VF*2, ..., VF*(NumVec-1), 1, VF+1, VF*2+1, ...>
+// E.g. For 2 interleaved vectors, if VF is 4, the mask is:
+// <0, 4, 1, 5, 2, 6, 3, 7>
+static Constant *getInterleavedMask(IRBuilder<> &Builder, unsigned VF,
+ unsigned NumVec) {
+ SmallVector<Constant *, 16> Mask;
+ for (unsigned i = 0; i < VF; i++)
+ for (unsigned j = 0; j < NumVec; j++)
+ Mask.push_back(Builder.getInt32(j * VF + i));
+
+ return ConstantVector::get(Mask);
+}
- Value *Start = Exp.expandCodeFor(PtrRtCheck->Starts[i], PtrArithTy, Loc);
- Value *End = Exp.expandCodeFor(PtrRtCheck->Ends[i], PtrArithTy, Loc);
- Starts.push_back(Start);
- Ends.push_back(End);
- }
- }
+// Get the strided mask starting from index \p Start.
+// I.e. <Start, Start + Stride, ..., Start + Stride*(VF-1)>
+static Constant *getStridedMask(IRBuilder<> &Builder, unsigned Start,
+ unsigned Stride, unsigned VF) {
+ SmallVector<Constant *, 16> Mask;
+ for (unsigned i = 0; i < VF; i++)
+ Mask.push_back(Builder.getInt32(Start + i * Stride));
- IRBuilder<> ChkBuilder(Loc);
- // Our instructions might fold to a constant.
- Value *MemoryRuntimeCheck = nullptr;
- for (unsigned i = 0; i < NumPointers; ++i) {
- for (unsigned j = i+1; j < NumPointers; ++j) {
- // No need to check if two readonly pointers intersect.
- if (!PtrRtCheck->IsWritePtr[i] && !PtrRtCheck->IsWritePtr[j])
- continue;
+ return ConstantVector::get(Mask);
+}
- // Only need to check pointers between two different dependency sets.
- if (PtrRtCheck->DependencySetId[i] == PtrRtCheck->DependencySetId[j])
- continue;
- // Only need to check pointers in the same alias set.
- if (PtrRtCheck->AliasSetId[i] != PtrRtCheck->AliasSetId[j])
- continue;
+// Get a mask of two parts: The first part consists of sequential integers
+// starting from 0, The second part consists of UNDEFs.
+// I.e. <0, 1, 2, ..., NumInt - 1, undef, ..., undef>
+static Constant *getSequentialMask(IRBuilder<> &Builder, unsigned NumInt,
+ unsigned NumUndef) {
+ SmallVector<Constant *, 16> Mask;
+ for (unsigned i = 0; i < NumInt; i++)
+ Mask.push_back(Builder.getInt32(i));
- unsigned AS0 = Starts[i]->getType()->getPointerAddressSpace();
- unsigned AS1 = Starts[j]->getType()->getPointerAddressSpace();
-
- assert((AS0 == Ends[j]->getType()->getPointerAddressSpace()) &&
- (AS1 == Ends[i]->getType()->getPointerAddressSpace()) &&
- "Trying to bounds check pointers with different address spaces");
-
- Type *PtrArithTy0 = Type::getInt8PtrTy(Ctx, AS0);
- Type *PtrArithTy1 = Type::getInt8PtrTy(Ctx, AS1);
-
- Value *Start0 = ChkBuilder.CreateBitCast(Starts[i], PtrArithTy0, "bc");
- Value *Start1 = ChkBuilder.CreateBitCast(Starts[j], PtrArithTy1, "bc");
- Value *End0 = ChkBuilder.CreateBitCast(Ends[i], PtrArithTy1, "bc");
- Value *End1 = ChkBuilder.CreateBitCast(Ends[j], PtrArithTy0, "bc");
-
- Value *Cmp0 = ChkBuilder.CreateICmpULE(Start0, End1, "bound0");
- FirstInst = getFirstInst(FirstInst, Cmp0, Loc);
- Value *Cmp1 = ChkBuilder.CreateICmpULE(Start1, End0, "bound1");
- FirstInst = getFirstInst(FirstInst, Cmp1, Loc);
- Value *IsConflict = ChkBuilder.CreateAnd(Cmp0, Cmp1, "found.conflict");
- FirstInst = getFirstInst(FirstInst, IsConflict, Loc);
- if (MemoryRuntimeCheck) {
- IsConflict = ChkBuilder.CreateOr(MemoryRuntimeCheck, IsConflict,
- "conflict.rdx");
- FirstInst = getFirstInst(FirstInst, IsConflict, Loc);
- }
- MemoryRuntimeCheck = IsConflict;
- }
- }
+ Constant *Undef = UndefValue::get(Builder.getInt32Ty());
+ for (unsigned i = 0; i < NumUndef; i++)
+ Mask.push_back(Undef);
- // We have to do this trickery because the IRBuilder might fold the check to a
- // constant expression in which case there is no Instruction anchored in a
- // the block.
- Instruction *Check = BinaryOperator::CreateAnd(MemoryRuntimeCheck,
- ConstantInt::getTrue(Ctx));
- ChkBuilder.Insert(Check, "memcheck.conflict");
- FirstInst = getFirstInst(FirstInst, Check, Loc);
- return std::make_pair(FirstInst, Check);
+ return ConstantVector::get(Mask);
}
-void InnerLoopVectorizer::createEmptyLoop() {
- /*
- In this function we generate a new loop. The new loop will contain
- the vectorized instructions while the old loop will continue to run the
- scalar remainder.
+// Concatenate two vectors with the same element type. The 2nd vector should
+// not have more elements than the 1st vector. If the 2nd vector has less
+// elements, extend it with UNDEFs.
+static Value *ConcatenateTwoVectors(IRBuilder<> &Builder, Value *V1,
+ Value *V2) {
+ VectorType *VecTy1 = dyn_cast<VectorType>(V1->getType());
+ VectorType *VecTy2 = dyn_cast<VectorType>(V2->getType());
+ assert(VecTy1 && VecTy2 &&
+ VecTy1->getScalarType() == VecTy2->getScalarType() &&
+ "Expect two vectors with the same element type");
+
+ unsigned NumElts1 = VecTy1->getNumElements();
+ unsigned NumElts2 = VecTy2->getNumElements();
+ assert(NumElts1 >= NumElts2 && "Unexpect the first vector has less elements");
+
+ if (NumElts1 > NumElts2) {
+ // Extend with UNDEFs.
+ Constant *ExtMask =
+ getSequentialMask(Builder, NumElts2, NumElts1 - NumElts2);
+ V2 = Builder.CreateShuffleVector(V2, UndefValue::get(VecTy2), ExtMask);
+ }
+
+ Constant *Mask = getSequentialMask(Builder, NumElts1 + NumElts2, 0);
+ return Builder.CreateShuffleVector(V1, V2, Mask);
+}
- [ ] <-- Back-edge taken count overflow check.
- / |
- / v
- | [ ] <-- vector loop bypass (may consist of multiple blocks).
- | / |
- | / v
- || [ ] <-- vector pre header.
- || |
- || v
- || [ ] \
- || [ ]_| <-- vector loop.
- || |
- | \ v
- | >[ ] <--- middle-block.
- | / |
- | / v
- -|- >[ ] <--- new preheader.
- | |
- | v
- | [ ] \
- | [ ]_| <-- old scalar loop to handle remainder.
- \ |
- \ v
- >[ ] <-- exit block.
- ...
- */
+// Concatenate vectors in the given list. All vectors have the same type.
+static Value *ConcatenateVectors(IRBuilder<> &Builder,
+ ArrayRef<Value *> InputList) {
+ unsigned NumVec = InputList.size();
+ assert(NumVec > 1 && "Should be at least two vectors");
+
+ SmallVector<Value *, 8> ResList;
+ ResList.append(InputList.begin(), InputList.end());
+ do {
+ SmallVector<Value *, 8> TmpList;
+ for (unsigned i = 0; i < NumVec - 1; i += 2) {
+ Value *V0 = ResList[i], *V1 = ResList[i + 1];
+ assert((V0->getType() == V1->getType() || i == NumVec - 2) &&
+ "Only the last vector may have a different type");
+
+ TmpList.push_back(ConcatenateTwoVectors(Builder, V0, V1));
+ }
- BasicBlock *OldBasicBlock = OrigLoop->getHeader();
- BasicBlock *BypassBlock = OrigLoop->getLoopPreheader();
- BasicBlock *ExitBlock = OrigLoop->getExitBlock();
- assert(BypassBlock && "Invalid loop structure");
- assert(ExitBlock && "Must have an exit block");
+ // Push the last vector if the total number of vectors is odd.
+ if (NumVec % 2 != 0)
+ TmpList.push_back(ResList[NumVec - 1]);
- // Some loops have a single integer induction variable, while other loops
- // don't. One example is c++ iterators that often have multiple pointer
- // induction variables. In the code below we also support a case where we
- // don't have a single induction variable.
- OldInduction = Legal->getInduction();
- Type *IdxTy = Legal->getWidestInductionType();
+ ResList = TmpList;
+ NumVec = ResList.size();
+ } while (NumVec > 1);
- // Find the loop boundaries.
- const SCEV *ExitCount = SE->getBackedgeTakenCount(OrigLoop);
- assert(ExitCount != SE->getCouldNotCompute() && "Invalid loop count");
+ return ResList[0];
+}
- // The exit count might have the type of i64 while the phi is i32. This can
- // happen if we have an induction variable that is sign extended before the
- // compare. The only way that we get a backedge taken count is that the
- // induction variable was signed and as such will not overflow. In such a case
- // truncation is legal.
- if (ExitCount->getType()->getPrimitiveSizeInBits() >
- IdxTy->getPrimitiveSizeInBits())
- ExitCount = SE->getTruncateOrNoop(ExitCount, IdxTy);
+// Try to vectorize the interleave group that \p Instr belongs to.
+//
+// E.g. Translate following interleaved load group (factor = 3):
+// for (i = 0; i < N; i+=3) {
+// R = Pic[i]; // Member of index 0
+// G = Pic[i+1]; // Member of index 1
+// B = Pic[i+2]; // Member of index 2
+// ... // do something to R, G, B
+// }
+// To:
+// %wide.vec = load <12 x i32> ; Read 4 tuples of R,G,B
+// %R.vec = shuffle %wide.vec, undef, <0, 3, 6, 9> ; R elements
+// %G.vec = shuffle %wide.vec, undef, <1, 4, 7, 10> ; G elements
+// %B.vec = shuffle %wide.vec, undef, <2, 5, 8, 11> ; B elements
+//
+// Or translate following interleaved store group (factor = 3):
+// for (i = 0; i < N; i+=3) {
+// ... do something to R, G, B
+// Pic[i] = R; // Member of index 0
+// Pic[i+1] = G; // Member of index 1
+// Pic[i+2] = B; // Member of index 2
+// }
+// To:
+// %R_G.vec = shuffle %R.vec, %G.vec, <0, 1, 2, ..., 7>
+// %B_U.vec = shuffle %B.vec, undef, <0, 1, 2, 3, u, u, u, u>
+// %interleaved.vec = shuffle %R_G.vec, %B_U.vec,
+// <0, 4, 8, 1, 5, 9, 2, 6, 10, 3, 7, 11> ; Interleave R,G,B elements
+// store <12 x i32> %interleaved.vec ; Write 4 tuples of R,G,B
+void InnerLoopVectorizer::vectorizeInterleaveGroup(Instruction *Instr) {
+ const InterleaveGroup *Group = Legal->getInterleavedAccessGroup(Instr);
+ assert(Group && "Fail to get an interleaved access group.");
+
+ // Skip if current instruction is not the insert position.
+ if (Instr != Group->getInsertPos())
+ return;
- const SCEV *BackedgeTakeCount = SE->getNoopOrZeroExtend(ExitCount, IdxTy);
- // Get the total trip count from the count by adding 1.
- ExitCount = SE->getAddExpr(BackedgeTakeCount,
- SE->getConstant(BackedgeTakeCount->getType(), 1));
+ LoadInst *LI = dyn_cast<LoadInst>(Instr);
+ StoreInst *SI = dyn_cast<StoreInst>(Instr);
+ Value *Ptr = LI ? LI->getPointerOperand() : SI->getPointerOperand();
- // Expand the trip count and place the new instructions in the preheader.
- // Notice that the pre-header does not change, only the loop body.
- SCEVExpander Exp(*SE, "induction");
+ // Prepare for the vector type of the interleaved load/store.
+ Type *ScalarTy = LI ? LI->getType() : SI->getValueOperand()->getType();
+ unsigned InterleaveFactor = Group->getFactor();
+ Type *VecTy = VectorType::get(ScalarTy, InterleaveFactor * VF);
+ Type *PtrTy = VecTy->getPointerTo(Ptr->getType()->getPointerAddressSpace());
+
+ // Prepare for the new pointers.
+ setDebugLocFromInst(Builder, Ptr);
+ VectorParts &PtrParts = getVectorValue(Ptr);
+ SmallVector<Value *, 2> NewPtrs;
+ unsigned Index = Group->getIndex(Instr);
+ for (unsigned Part = 0; Part < UF; Part++) {
+ // Extract the pointer for current instruction from the pointer vector. A
+ // reverse access uses the pointer in the last lane.
+ Value *NewPtr = Builder.CreateExtractElement(
+ PtrParts[Part],
+ Group->isReverse() ? Builder.getInt32(VF - 1) : Builder.getInt32(0));
+
+ // Notice current instruction could be any index. Need to adjust the address
+ // to the member of index 0.
+ //
+ // E.g. a = A[i+1]; // Member of index 1 (Current instruction)
+ // b = A[i]; // Member of index 0
+ // Current pointer is pointed to A[i+1], adjust it to A[i].
+ //
+ // E.g. A[i+1] = a; // Member of index 1
+ // A[i] = b; // Member of index 0
+ // A[i+2] = c; // Member of index 2 (Current instruction)
+ // Current pointer is pointed to A[i+2], adjust it to A[i].
+ NewPtr = Builder.CreateGEP(NewPtr, Builder.getInt32(-Index));
- // We need to test whether the backedge-taken count is uint##_max. Adding one
- // to it will cause overflow and an incorrect loop trip count in the vector
- // body. In case of overflow we want to directly jump to the scalar remainder
- // loop.
- Value *BackedgeCount =
- Exp.expandCodeFor(BackedgeTakeCount, BackedgeTakeCount->getType(),
- BypassBlock->getTerminator());
- if (BackedgeCount->getType()->isPointerTy())
- BackedgeCount = CastInst::CreatePointerCast(BackedgeCount, IdxTy,
- "backedge.ptrcnt.to.int",
- BypassBlock->getTerminator());
- Instruction *CheckBCOverflow =
- CmpInst::Create(Instruction::ICmp, CmpInst::ICMP_EQ, BackedgeCount,
- Constant::getAllOnesValue(BackedgeCount->getType()),
- "backedge.overflow", BypassBlock->getTerminator());
+ // Cast to the vector pointer type.
+ NewPtrs.push_back(Builder.CreateBitCast(NewPtr, PtrTy));
+ }
- // The loop index does not have to start at Zero. Find the original start
- // value from the induction PHI node. If we don't have an induction variable
- // then we know that it starts at zero.
- Builder.SetInsertPoint(BypassBlock->getTerminator());
- Value *StartIdx = ExtendedIdx = OldInduction ?
- Builder.CreateZExt(OldInduction->getIncomingValueForBlock(BypassBlock),
- IdxTy):
- ConstantInt::get(IdxTy, 0);
-
- // We need an instruction to anchor the overflow check on. StartIdx needs to
- // be defined before the overflow check branch. Because the scalar preheader
- // is going to merge the start index and so the overflow branch block needs to
- // contain a definition of the start index.
- Instruction *OverflowCheckAnchor = BinaryOperator::CreateAdd(
- StartIdx, ConstantInt::get(IdxTy, 0), "overflow.check.anchor",
- BypassBlock->getTerminator());
+ setDebugLocFromInst(Builder, Instr);
+ Value *UndefVec = UndefValue::get(VecTy);
- // Count holds the overall loop count (N).
- Value *Count = Exp.expandCodeFor(ExitCount, ExitCount->getType(),
- BypassBlock->getTerminator());
+ // Vectorize the interleaved load group.
+ if (LI) {
+ for (unsigned Part = 0; Part < UF; Part++) {
+ Instruction *NewLoadInstr = Builder.CreateAlignedLoad(
+ NewPtrs[Part], Group->getAlignment(), "wide.vec");
- LoopBypassBlocks.push_back(BypassBlock);
+ for (unsigned i = 0; i < InterleaveFactor; i++) {
+ Instruction *Member = Group->getMember(i);
- // Split the single block loop into the two loop structure described above.
- BasicBlock *VectorPH =
- BypassBlock->splitBasicBlock(BypassBlock->getTerminator(), "vector.ph");
- BasicBlock *VecBody =
- VectorPH->splitBasicBlock(VectorPH->getTerminator(), "vector.body");
- BasicBlock *MiddleBlock =
- VecBody->splitBasicBlock(VecBody->getTerminator(), "middle.block");
- BasicBlock *ScalarPH =
- MiddleBlock->splitBasicBlock(MiddleBlock->getTerminator(), "scalar.ph");
+ // Skip the gaps in the group.
+ if (!Member)
+ continue;
- // Create and register the new vector loop.
- Loop* Lp = new Loop();
- Loop *ParentLoop = OrigLoop->getParentLoop();
+ Constant *StrideMask = getStridedMask(Builder, i, InterleaveFactor, VF);
+ Value *StridedVec = Builder.CreateShuffleVector(
+ NewLoadInstr, UndefVec, StrideMask, "strided.vec");
- // Insert the new loop into the loop nest and register the new basic blocks
- // before calling any utilities such as SCEV that require valid LoopInfo.
- if (ParentLoop) {
- ParentLoop->addChildLoop(Lp);
- ParentLoop->addBasicBlockToLoop(ScalarPH, *LI);
- ParentLoop->addBasicBlockToLoop(VectorPH, *LI);
- ParentLoop->addBasicBlockToLoop(MiddleBlock, *LI);
- } else {
- LI->addTopLevelLoop(Lp);
+ // If this member has different type, cast the result type.
+ if (Member->getType() != ScalarTy) {
+ VectorType *OtherVTy = VectorType::get(Member->getType(), VF);
+ StridedVec = Builder.CreateBitOrPointerCast(StridedVec, OtherVTy);
+ }
+
+ VectorParts &Entry = WidenMap.get(Member);
+ Entry[Part] =
+ Group->isReverse() ? reverseVector(StridedVec) : StridedVec;
+ }
+
+ propagateMetadata(NewLoadInstr, Instr);
+ }
+ return;
}
- Lp->addBasicBlockToLoop(VecBody, *LI);
- // Use this IR builder to create the loop instructions (Phi, Br, Cmp)
- // inside the loop.
- Builder.SetInsertPoint(VecBody->getFirstNonPHI());
+ // The sub vector type for current instruction.
+ VectorType *SubVT = VectorType::get(ScalarTy, VF);
- // Generate the induction variable.
- setDebugLocFromInst(Builder, getDebugLocFromInstOrOperands(OldInduction));
- Induction = Builder.CreatePHI(IdxTy, 2, "index");
- // The loop step is equal to the vectorization factor (num of SIMD elements)
- // times the unroll factor (num of SIMD instructions).
- Constant *Step = ConstantInt::get(IdxTy, VF * UF);
+ // Vectorize the interleaved store group.
+ for (unsigned Part = 0; Part < UF; Part++) {
+ // Collect the stored vector from each member.
+ SmallVector<Value *, 4> StoredVecs;
+ for (unsigned i = 0; i < InterleaveFactor; i++) {
+ // Interleaved store group doesn't allow a gap, so each index has a member
+ Instruction *Member = Group->getMember(i);
+ assert(Member && "Fail to get a member from an interleaved store group");
- // This is the IR builder that we use to add all of the logic for bypassing
- // the new vector loop.
- IRBuilder<> BypassBuilder(BypassBlock->getTerminator());
- setDebugLocFromInst(BypassBuilder,
- getDebugLocFromInstOrOperands(OldInduction));
+ Value *StoredVec =
+ getVectorValue(dyn_cast<StoreInst>(Member)->getValueOperand())[Part];
+ if (Group->isReverse())
+ StoredVec = reverseVector(StoredVec);
- // We may need to extend the index in case there is a type mismatch.
- // We know that the count starts at zero and does not overflow.
- if (Count->getType() != IdxTy) {
- // The exit count can be of pointer type. Convert it to the correct
- // integer type.
- if (ExitCount->getType()->isPointerTy())
- Count = BypassBuilder.CreatePointerCast(Count, IdxTy, "ptrcnt.to.int");
- else
- Count = BypassBuilder.CreateZExtOrTrunc(Count, IdxTy, "cnt.cast");
- }
-
- // Add the start index to the loop count to get the new end index.
- Value *IdxEnd = BypassBuilder.CreateAdd(Count, StartIdx, "end.idx");
+ // If this member has different type, cast it to an unified type.
+ if (StoredVec->getType() != SubVT)
+ StoredVec = Builder.CreateBitOrPointerCast(StoredVec, SubVT);
- // Now we need to generate the expression for N - (N % VF), which is
- // the part that the vectorized body will execute.
- Value *R = BypassBuilder.CreateURem(Count, Step, "n.mod.vf");
- Value *CountRoundDown = BypassBuilder.CreateSub(Count, R, "n.vec");
- Value *IdxEndRoundDown = BypassBuilder.CreateAdd(CountRoundDown, StartIdx,
- "end.idx.rnd.down");
+ StoredVecs.push_back(StoredVec);
+ }
- // Now, compare the new count to zero. If it is zero skip the vector loop and
- // jump to the scalar loop.
- Value *Cmp =
- BypassBuilder.CreateICmpEQ(IdxEndRoundDown, StartIdx, "cmp.zero");
+ // Concatenate all vectors into a wide vector.
+ Value *WideVec = ConcatenateVectors(Builder, StoredVecs);
- BasicBlock *LastBypassBlock = BypassBlock;
+ // Interleave the elements in the wide vector.
+ Constant *IMask = getInterleavedMask(Builder, VF, InterleaveFactor);
+ Value *IVec = Builder.CreateShuffleVector(WideVec, UndefVec, IMask,
+ "interleaved.vec");
- // Generate code to check that the loops trip count that we computed by adding
- // one to the backedge-taken count will not overflow.
- {
- auto PastOverflowCheck =
- std::next(BasicBlock::iterator(OverflowCheckAnchor));
- BasicBlock *CheckBlock =
- LastBypassBlock->splitBasicBlock(PastOverflowCheck, "overflow.checked");
- if (ParentLoop)
- ParentLoop->addBasicBlockToLoop(CheckBlock, *LI);
- LoopBypassBlocks.push_back(CheckBlock);
- Instruction *OldTerm = LastBypassBlock->getTerminator();
- BranchInst::Create(ScalarPH, CheckBlock, CheckBCOverflow, OldTerm);
- OldTerm->eraseFromParent();
- LastBypassBlock = CheckBlock;
+ Instruction *NewStoreInstr =
+ Builder.CreateAlignedStore(IVec, NewPtrs[Part], Group->getAlignment());
+ propagateMetadata(NewStoreInstr, Instr);
}
+}
- // Generate the code to check that the strides we assumed to be one are really
- // one. We want the new basic block to start at the first instruction in a
- // sequence of instructions that form a check.
- Instruction *StrideCheck;
- Instruction *FirstCheckInst;
- std::tie(FirstCheckInst, StrideCheck) =
- addStrideCheck(LastBypassBlock->getTerminator());
- if (StrideCheck) {
- // Create a new block containing the stride check.
- BasicBlock *CheckBlock =
- LastBypassBlock->splitBasicBlock(FirstCheckInst, "vector.stridecheck");
- if (ParentLoop)
- ParentLoop->addBasicBlockToLoop(CheckBlock, *LI);
- LoopBypassBlocks.push_back(CheckBlock);
+void InnerLoopVectorizer::vectorizeMemoryInstruction(Instruction *Instr) {
+ // Attempt to issue a wide load.
+ LoadInst *LI = dyn_cast<LoadInst>(Instr);
+ StoreInst *SI = dyn_cast<StoreInst>(Instr);
- // Replace the branch into the memory check block with a conditional branch
- // for the "few elements case".
- Instruction *OldTerm = LastBypassBlock->getTerminator();
- BranchInst::Create(MiddleBlock, CheckBlock, Cmp, OldTerm);
- OldTerm->eraseFromParent();
+ assert((LI || SI) && "Invalid Load/Store instruction");
- Cmp = StrideCheck;
- LastBypassBlock = CheckBlock;
- }
+ // Try to vectorize the interleave group if this access is interleaved.
+ if (Legal->isAccessInterleaved(Instr))
+ return vectorizeInterleaveGroup(Instr);
- // Generate the code that checks in runtime if arrays overlap. We put the
- // checks into a separate block to make the more common case of few elements
- // faster.
- Instruction *MemRuntimeCheck;
- std::tie(FirstCheckInst, MemRuntimeCheck) =
- addRuntimeCheck(LastBypassBlock->getTerminator());
- if (MemRuntimeCheck) {
- // Create a new block containing the memory check.
- BasicBlock *CheckBlock =
- LastBypassBlock->splitBasicBlock(MemRuntimeCheck, "vector.memcheck");
- if (ParentLoop)
- ParentLoop->addBasicBlockToLoop(CheckBlock, *LI);
- LoopBypassBlocks.push_back(CheckBlock);
+ Type *ScalarDataTy = LI ? LI->getType() : SI->getValueOperand()->getType();
+ Type *DataTy = VectorType::get(ScalarDataTy, VF);
+ Value *Ptr = LI ? LI->getPointerOperand() : SI->getPointerOperand();
+ unsigned Alignment = LI ? LI->getAlignment() : SI->getAlignment();
+ // An alignment of 0 means target abi alignment. We need to use the scalar's
+ // target abi alignment in such a case.
+ const DataLayout &DL = Instr->getModule()->getDataLayout();
+ if (!Alignment)
+ Alignment = DL.getABITypeAlignment(ScalarDataTy);
+ unsigned AddressSpace = Ptr->getType()->getPointerAddressSpace();
+ unsigned ScalarAllocatedSize = DL.getTypeAllocSize(ScalarDataTy);
+ unsigned VectorElementSize = DL.getTypeStoreSize(DataTy) / VF;
- // Replace the branch into the memory check block with a conditional branch
- // for the "few elements case".
- Instruction *OldTerm = LastBypassBlock->getTerminator();
- BranchInst::Create(MiddleBlock, CheckBlock, Cmp, OldTerm);
- OldTerm->eraseFromParent();
+ if (SI && Legal->blockNeedsPredication(SI->getParent()) &&
+ !Legal->isMaskRequired(SI))
+ return scalarizeInstruction(Instr, true);
- Cmp = MemRuntimeCheck;
- LastBypassBlock = CheckBlock;
- }
+ if (ScalarAllocatedSize != VectorElementSize)
+ return scalarizeInstruction(Instr);
- LastBypassBlock->getTerminator()->eraseFromParent();
- BranchInst::Create(MiddleBlock, VectorPH, Cmp,
- LastBypassBlock);
+ // If the pointer is loop invariant or if it is non-consecutive,
+ // scalarize the load.
+ int ConsecutiveStride = Legal->isConsecutivePtr(Ptr);
+ bool Reverse = ConsecutiveStride < 0;
+ bool UniformLoad = LI && Legal->isUniform(Ptr);
+ if (!ConsecutiveStride || UniformLoad)
+ return scalarizeInstruction(Instr);
- // We are going to resume the execution of the scalar loop.
- // Go over all of the induction variables that we found and fix the
- // PHIs that are left in the scalar version of the loop.
- // The starting values of PHI nodes depend on the counter of the last
- // iteration in the vectorized loop.
- // If we come from a bypass edge then we need to start from the original
- // start value.
+ Constant *Zero = Builder.getInt32(0);
+ VectorParts &Entry = WidenMap.get(Instr);
- // This variable saves the new starting index for the scalar loop.
- PHINode *ResumeIndex = nullptr;
- LoopVectorizationLegality::InductionList::iterator I, E;
- LoopVectorizationLegality::InductionList *List = Legal->getInductionVars();
- // Set builder to point to last bypass block.
- BypassBuilder.SetInsertPoint(LoopBypassBlocks.back()->getTerminator());
- for (I = List->begin(), E = List->end(); I != E; ++I) {
- PHINode *OrigPhi = I->first;
- LoopVectorizationLegality::InductionInfo II = I->second;
+ // Handle consecutive loads/stores.
+ GetElementPtrInst *Gep = dyn_cast<GetElementPtrInst>(Ptr);
+ if (Gep && Legal->isInductionVariable(Gep->getPointerOperand())) {
+ setDebugLocFromInst(Builder, Gep);
+ Value *PtrOperand = Gep->getPointerOperand();
+ Value *FirstBasePtr = getVectorValue(PtrOperand)[0];
+ FirstBasePtr = Builder.CreateExtractElement(FirstBasePtr, Zero);
- Type *ResumeValTy = (OrigPhi == OldInduction) ? IdxTy : OrigPhi->getType();
- PHINode *ResumeVal = PHINode::Create(ResumeValTy, 2, "resume.val",
- MiddleBlock->getTerminator());
- // We might have extended the type of the induction variable but we need a
- // truncated version for the scalar loop.
- PHINode *TruncResumeVal = (OrigPhi == OldInduction) ?
- PHINode::Create(OrigPhi->getType(), 2, "trunc.resume.val",
- MiddleBlock->getTerminator()) : nullptr;
+ // Create the new GEP with the new induction variable.
+ GetElementPtrInst *Gep2 = cast<GetElementPtrInst>(Gep->clone());
+ Gep2->setOperand(0, FirstBasePtr);
+ Gep2->setName("gep.indvar.base");
+ Ptr = Builder.Insert(Gep2);
+ } else if (Gep) {
+ setDebugLocFromInst(Builder, Gep);
+ assert(SE->isLoopInvariant(SE->getSCEV(Gep->getPointerOperand()),
+ OrigLoop) && "Base ptr must be invariant");
- // Create phi nodes to merge from the backedge-taken check block.
- PHINode *BCResumeVal = PHINode::Create(ResumeValTy, 3, "bc.resume.val",
- ScalarPH->getTerminator());
- BCResumeVal->addIncoming(ResumeVal, MiddleBlock);
+ // The last index does not have to be the induction. It can be
+ // consecutive and be a function of the index. For example A[I+1];
+ unsigned NumOperands = Gep->getNumOperands();
+ unsigned InductionOperand = getGEPInductionOperand(Gep);
+ // Create the new GEP with the new induction variable.
+ GetElementPtrInst *Gep2 = cast<GetElementPtrInst>(Gep->clone());
- PHINode *BCTruncResumeVal = nullptr;
- if (OrigPhi == OldInduction) {
- BCTruncResumeVal =
- PHINode::Create(OrigPhi->getType(), 2, "bc.trunc.resume.val",
- ScalarPH->getTerminator());
- BCTruncResumeVal->addIncoming(TruncResumeVal, MiddleBlock);
- }
+ for (unsigned i = 0; i < NumOperands; ++i) {
+ Value *GepOperand = Gep->getOperand(i);
+ Instruction *GepOperandInst = dyn_cast<Instruction>(GepOperand);
- Value *EndValue = nullptr;
- switch (II.IK) {
- case LoopVectorizationLegality::IK_NoInduction:
- llvm_unreachable("Unknown induction");
- case LoopVectorizationLegality::IK_IntInduction: {
- // Handle the integer induction counter.
- assert(OrigPhi->getType()->isIntegerTy() && "Invalid type");
+ // Update last index or loop invariant instruction anchored in loop.
+ if (i == InductionOperand ||
+ (GepOperandInst && OrigLoop->contains(GepOperandInst))) {
+ assert((i == InductionOperand ||
+ SE->isLoopInvariant(SE->getSCEV(GepOperandInst), OrigLoop)) &&
+ "Must be last index or loop invariant");
- // We have the canonical induction variable.
- if (OrigPhi == OldInduction) {
- // Create a truncated version of the resume value for the scalar loop,
- // we might have promoted the type to a larger width.
- EndValue =
- BypassBuilder.CreateTrunc(IdxEndRoundDown, OrigPhi->getType());
- // The new PHI merges the original incoming value, in case of a bypass,
- // or the value at the end of the vectorized loop.
- for (unsigned I = 1, E = LoopBypassBlocks.size(); I != E; ++I)
- TruncResumeVal->addIncoming(II.StartValue, LoopBypassBlocks[I]);
- TruncResumeVal->addIncoming(EndValue, VecBody);
+ VectorParts &GEPParts = getVectorValue(GepOperand);
+ Value *Index = GEPParts[0];
+ Index = Builder.CreateExtractElement(Index, Zero);
+ Gep2->setOperand(i, Index);
+ Gep2->setName("gep.indvar.idx");
+ }
+ }
+ Ptr = Builder.Insert(Gep2);
+ } else {
+ // Use the induction element ptr.
+ assert(isa<PHINode>(Ptr) && "Invalid induction ptr");
+ setDebugLocFromInst(Builder, Ptr);
+ VectorParts &PtrVal = getVectorValue(Ptr);
+ Ptr = Builder.CreateExtractElement(PtrVal[0], Zero);
+ }
- BCTruncResumeVal->addIncoming(II.StartValue, LoopBypassBlocks[0]);
+ VectorParts Mask = createBlockInMask(Instr->getParent());
+ // Handle Stores:
+ if (SI) {
+ assert(!Legal->isUniform(SI->getPointerOperand()) &&
+ "We do not allow storing to uniform addresses");
+ setDebugLocFromInst(Builder, SI);
+ // We don't want to update the value in the map as it might be used in
+ // another expression. So don't use a reference type for "StoredVal".
+ VectorParts StoredVal = getVectorValue(SI->getValueOperand());
+
+ for (unsigned Part = 0; Part < UF; ++Part) {
+ // Calculate the pointer for the specific unroll-part.
+ Value *PartPtr =
+ Builder.CreateGEP(nullptr, Ptr, Builder.getInt32(Part * VF));
- // We know what the end value is.
- EndValue = IdxEndRoundDown;
- // We also know which PHI node holds it.
- ResumeIndex = ResumeVal;
- break;
+ if (Reverse) {
+ // If we store to reverse consecutive memory locations, then we need
+ // to reverse the order of elements in the stored value.
+ StoredVal[Part] = reverseVector(StoredVal[Part]);
+ // If the address is consecutive but reversed, then the
+ // wide store needs to start at the last vector element.
+ PartPtr = Builder.CreateGEP(nullptr, Ptr, Builder.getInt32(-Part * VF));
+ PartPtr = Builder.CreateGEP(nullptr, PartPtr, Builder.getInt32(1 - VF));
+ Mask[Part] = reverseVector(Mask[Part]);
}
- // Not the canonical induction variable - add the vector loop count to the
- // start value.
- Value *CRD = BypassBuilder.CreateSExtOrTrunc(CountRoundDown,
- II.StartValue->getType(),
- "cast.crd");
- EndValue = II.transform(BypassBuilder, CRD);
- EndValue->setName("ind.end");
- break;
- }
- case LoopVectorizationLegality::IK_PtrInduction: {
- EndValue = II.transform(BypassBuilder, CountRoundDown);
- EndValue->setName("ptr.ind.end");
- break;
- }
- }// end of case
+ Value *VecPtr = Builder.CreateBitCast(PartPtr,
+ DataTy->getPointerTo(AddressSpace));
- // The new PHI merges the original incoming value, in case of a bypass,
- // or the value at the end of the vectorized loop.
- for (unsigned I = 1, E = LoopBypassBlocks.size(); I != E; ++I) {
- if (OrigPhi == OldInduction)
- ResumeVal->addIncoming(StartIdx, LoopBypassBlocks[I]);
- else
- ResumeVal->addIncoming(II.StartValue, LoopBypassBlocks[I]);
+ Instruction *NewSI;
+ if (Legal->isMaskRequired(SI))
+ NewSI = Builder.CreateMaskedStore(StoredVal[Part], VecPtr, Alignment,
+ Mask[Part]);
+ else
+ NewSI = Builder.CreateAlignedStore(StoredVal[Part], VecPtr, Alignment);
+ propagateMetadata(NewSI, SI);
}
- ResumeVal->addIncoming(EndValue, VecBody);
+ return;
+ }
- // Fix the scalar body counter (PHI node).
- unsigned BlockIdx = OrigPhi->getBasicBlockIndex(ScalarPH);
+ // Handle loads.
+ assert(LI && "Must have a load instruction");
+ setDebugLocFromInst(Builder, LI);
+ for (unsigned Part = 0; Part < UF; ++Part) {
+ // Calculate the pointer for the specific unroll-part.
+ Value *PartPtr =
+ Builder.CreateGEP(nullptr, Ptr, Builder.getInt32(Part * VF));
- // The old induction's phi node in the scalar body needs the truncated
- // value.
- if (OrigPhi == OldInduction) {
- BCResumeVal->addIncoming(StartIdx, LoopBypassBlocks[0]);
- OrigPhi->setIncomingValue(BlockIdx, BCTruncResumeVal);
- } else {
- BCResumeVal->addIncoming(II.StartValue, LoopBypassBlocks[0]);
- OrigPhi->setIncomingValue(BlockIdx, BCResumeVal);
+ if (Reverse) {
+ // If the address is consecutive but reversed, then the
+ // wide load needs to start at the last vector element.
+ PartPtr = Builder.CreateGEP(nullptr, Ptr, Builder.getInt32(-Part * VF));
+ PartPtr = Builder.CreateGEP(nullptr, PartPtr, Builder.getInt32(1 - VF));
+ Mask[Part] = reverseVector(Mask[Part]);
}
- }
- // If we are generating a new induction variable then we also need to
- // generate the code that calculates the exit value. This value is not
- // simply the end of the counter because we may skip the vectorized body
- // in case of a runtime check.
- if (!OldInduction){
- assert(!ResumeIndex && "Unexpected resume value found");
- ResumeIndex = PHINode::Create(IdxTy, 2, "new.indc.resume.val",
- MiddleBlock->getTerminator());
- for (unsigned I = 1, E = LoopBypassBlocks.size(); I != E; ++I)
- ResumeIndex->addIncoming(StartIdx, LoopBypassBlocks[I]);
- ResumeIndex->addIncoming(IdxEndRoundDown, VecBody);
+ Instruction* NewLI;
+ Value *VecPtr = Builder.CreateBitCast(PartPtr,
+ DataTy->getPointerTo(AddressSpace));
+ if (Legal->isMaskRequired(LI))
+ NewLI = Builder.CreateMaskedLoad(VecPtr, Alignment, Mask[Part],
+ UndefValue::get(DataTy),
+ "wide.masked.load");
+ else
+ NewLI = Builder.CreateAlignedLoad(VecPtr, Alignment, "wide.load");
+ propagateMetadata(NewLI, LI);
+ Entry[Part] = Reverse ? reverseVector(NewLI) : NewLI;
}
+}
- // Make sure that we found the index where scalar loop needs to continue.
- assert(ResumeIndex && ResumeIndex->getType()->isIntegerTy() &&
- "Invalid resume Index");
-
- // Add a check in the middle block to see if we have completed
- // all of the iterations in the first vector loop.
- // If (N - N%VF) == N, then we *don't* need to run the remainder.
- Value *CmpN = CmpInst::Create(Instruction::ICmp, CmpInst::ICMP_EQ, IdxEnd,
- ResumeIndex, "cmp.n",
- MiddleBlock->getTerminator());
-
- BranchInst::Create(ExitBlock, ScalarPH, CmpN, MiddleBlock->getTerminator());
- // Remove the old terminator.
- MiddleBlock->getTerminator()->eraseFromParent();
-
- // Create i+1 and fill the PHINode.
- Value *NextIdx = Builder.CreateAdd(Induction, Step, "index.next");
- Induction->addIncoming(StartIdx, VectorPH);
- Induction->addIncoming(NextIdx, VecBody);
- // Create the compare.
- Value *ICmp = Builder.CreateICmpEQ(NextIdx, IdxEndRoundDown);
- Builder.CreateCondBr(ICmp, MiddleBlock, VecBody);
+void InnerLoopVectorizer::scalarizeInstruction(Instruction *Instr, bool IfPredicateStore) {
+ assert(!Instr->getType()->isAggregateType() && "Can't handle vectors");
+ // Holds vector parameters or scalars, in case of uniform vals.
+ SmallVector<VectorParts, 4> Params;
- // Now we have two terminators. Remove the old one from the block.
- VecBody->getTerminator()->eraseFromParent();
+ setDebugLocFromInst(Builder, Instr);
- // Get ready to start creating new instructions into the vectorized body.
- Builder.SetInsertPoint(VecBody->getFirstInsertionPt());
+ // Find all of the vectorized parameters.
+ for (unsigned op = 0, e = Instr->getNumOperands(); op != e; ++op) {
+ Value *SrcOp = Instr->getOperand(op);
- // Save the state.
- LoopVectorPreHeader = VectorPH;
- LoopScalarPreHeader = ScalarPH;
- LoopMiddleBlock = MiddleBlock;
- LoopExitBlock = ExitBlock;
- LoopVectorBody.push_back(VecBody);
- LoopScalarBody = OldBasicBlock;
+ // If we are accessing the old induction variable, use the new one.
+ if (SrcOp == OldInduction) {
+ Params.push_back(getVectorValue(SrcOp));
+ continue;
+ }
- LoopVectorizeHints Hints(Lp, true);
- Hints.setAlreadyVectorized();
-}
+ // Try using previously calculated values.
+ Instruction *SrcInst = dyn_cast<Instruction>(SrcOp);
-/// This function returns the identity element (or neutral element) for
-/// the operation K.
-Constant*
-LoopVectorizationLegality::getReductionIdentity(ReductionKind K, Type *Tp) {
- switch (K) {
- case RK_IntegerXor:
- case RK_IntegerAdd:
- case RK_IntegerOr:
- // Adding, Xoring, Oring zero to a number does not change it.
- return ConstantInt::get(Tp, 0);
- case RK_IntegerMult:
- // Multiplying a number by 1 does not change it.
- return ConstantInt::get(Tp, 1);
- case RK_IntegerAnd:
- // AND-ing a number with an all-1 value does not change it.
- return ConstantInt::get(Tp, -1, true);
- case RK_FloatMult:
- // Multiplying a number by 1 does not change it.
- return ConstantFP::get(Tp, 1.0L);
- case RK_FloatAdd:
- // Adding zero to a number does not change it.
- return ConstantFP::get(Tp, 0.0L);
- default:
- llvm_unreachable("Unknown reduction kind");
+ // If the src is an instruction that appeared earlier in the basic block,
+ // then it should already be vectorized.
+ if (SrcInst && OrigLoop->contains(SrcInst)) {
+ assert(WidenMap.has(SrcInst) && "Source operand is unavailable");
+ // The parameter is a vector value from earlier.
+ Params.push_back(WidenMap.get(SrcInst));
+ } else {
+ // The parameter is a scalar from outside the loop. Maybe even a constant.
+ VectorParts Scalars;
+ Scalars.append(UF, SrcOp);
+ Params.push_back(Scalars);
+ }
}
-}
-/// This function translates the reduction kind to an LLVM binary operator.
-static unsigned
-getReductionBinOp(LoopVectorizationLegality::ReductionKind Kind) {
- switch (Kind) {
- case LoopVectorizationLegality::RK_IntegerAdd:
- return Instruction::Add;
- case LoopVectorizationLegality::RK_IntegerMult:
- return Instruction::Mul;
- case LoopVectorizationLegality::RK_IntegerOr:
- return Instruction::Or;
- case LoopVectorizationLegality::RK_IntegerAnd:
- return Instruction::And;
- case LoopVectorizationLegality::RK_IntegerXor:
- return Instruction::Xor;
- case LoopVectorizationLegality::RK_FloatMult:
- return Instruction::FMul;
- case LoopVectorizationLegality::RK_FloatAdd:
- return Instruction::FAdd;
- case LoopVectorizationLegality::RK_IntegerMinMax:
- return Instruction::ICmp;
- case LoopVectorizationLegality::RK_FloatMinMax:
- return Instruction::FCmp;
- default:
- llvm_unreachable("Unknown reduction operation");
- }
-}
+ assert(Params.size() == Instr->getNumOperands() &&
+ "Invalid number of operands");
-Value *createMinMaxOp(IRBuilder<> &Builder,
- LoopVectorizationLegality::MinMaxReductionKind RK,
- Value *Left,
- Value *Right) {
- CmpInst::Predicate P = CmpInst::ICMP_NE;
- switch (RK) {
- default:
- llvm_unreachable("Unknown min/max reduction kind");
- case LoopVectorizationLegality::MRK_UIntMin:
- P = CmpInst::ICMP_ULT;
- break;
- case LoopVectorizationLegality::MRK_UIntMax:
- P = CmpInst::ICMP_UGT;
- break;
- case LoopVectorizationLegality::MRK_SIntMin:
- P = CmpInst::ICMP_SLT;
- break;
- case LoopVectorizationLegality::MRK_SIntMax:
- P = CmpInst::ICMP_SGT;
- break;
- case LoopVectorizationLegality::MRK_FloatMin:
- P = CmpInst::FCMP_OLT;
- break;
- case LoopVectorizationLegality::MRK_FloatMax:
- P = CmpInst::FCMP_OGT;
- break;
- }
-
- Value *Cmp;
- if (RK == LoopVectorizationLegality::MRK_FloatMin ||
- RK == LoopVectorizationLegality::MRK_FloatMax)
- Cmp = Builder.CreateFCmp(P, Left, Right, "rdx.minmax.cmp");
- else
- Cmp = Builder.CreateICmp(P, Left, Right, "rdx.minmax.cmp");
+ // Does this instruction return a value ?
+ bool IsVoidRetTy = Instr->getType()->isVoidTy();
- Value *Select = Builder.CreateSelect(Cmp, Left, Right, "rdx.minmax.select");
- return Select;
-}
+ Value *UndefVec = IsVoidRetTy ? nullptr :
+ UndefValue::get(VectorType::get(Instr->getType(), VF));
+ // Create a new entry in the WidenMap and initialize it to Undef or Null.
+ VectorParts &VecResults = WidenMap.splat(Instr, UndefVec);
-namespace {
-struct CSEDenseMapInfo {
- static bool canHandle(Instruction *I) {
- return isa<InsertElementInst>(I) || isa<ExtractElementInst>(I) ||
- isa<ShuffleVectorInst>(I) || isa<GetElementPtrInst>(I);
- }
- static inline Instruction *getEmptyKey() {
- return DenseMapInfo<Instruction *>::getEmptyKey();
- }
- static inline Instruction *getTombstoneKey() {
- return DenseMapInfo<Instruction *>::getTombstoneKey();
- }
- static unsigned getHashValue(Instruction *I) {
- assert(canHandle(I) && "Unknown instruction!");
- return hash_combine(I->getOpcode(), hash_combine_range(I->value_op_begin(),
- I->value_op_end()));
- }
- static bool isEqual(Instruction *LHS, Instruction *RHS) {
- if (LHS == getEmptyKey() || RHS == getEmptyKey() ||
- LHS == getTombstoneKey() || RHS == getTombstoneKey())
- return LHS == RHS;
- return LHS->isIdenticalTo(RHS);
- }
-};
-}
+ Instruction *InsertPt = Builder.GetInsertPoint();
+ BasicBlock *IfBlock = Builder.GetInsertBlock();
+ BasicBlock *CondBlock = nullptr;
-/// \brief Check whether this block is a predicated block.
-/// Due to if predication of stores we might create a sequence of "if(pred) a[i]
-/// = ...; " blocks. We start with one vectorized basic block. For every
-/// conditional block we split this vectorized block. Therefore, every second
-/// block will be a predicated one.
-static bool isPredicatedBlock(unsigned BlockNum) {
- return BlockNum % 2;
-}
+ VectorParts Cond;
+ Loop *VectorLp = nullptr;
+ if (IfPredicateStore) {
+ assert(Instr->getParent()->getSinglePredecessor() &&
+ "Only support single predecessor blocks");
+ Cond = createEdgeMask(Instr->getParent()->getSinglePredecessor(),
+ Instr->getParent());
+ VectorLp = LI->getLoopFor(IfBlock);
+ assert(VectorLp && "Must have a loop for this block");
+ }
-///\brief Perform cse of induction variable instructions.
-static void cse(SmallVector<BasicBlock *, 4> &BBs) {
- // Perform simple cse.
- SmallDenseMap<Instruction *, Instruction *, 4, CSEDenseMapInfo> CSEMap;
- for (unsigned i = 0, e = BBs.size(); i != e; ++i) {
- BasicBlock *BB = BBs[i];
- for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E;) {
- Instruction *In = I++;
+ // For each vector unroll 'part':
+ for (unsigned Part = 0; Part < UF; ++Part) {
+ // For each scalar that we create:
+ for (unsigned Width = 0; Width < VF; ++Width) {
- if (!CSEDenseMapInfo::canHandle(In))
- continue;
+ // Start if-block.
+ Value *Cmp = nullptr;
+ if (IfPredicateStore) {
+ Cmp = Builder.CreateExtractElement(Cond[Part], Builder.getInt32(Width));
+ Cmp = Builder.CreateICmp(ICmpInst::ICMP_EQ, Cmp, ConstantInt::get(Cmp->getType(), 1));
+ CondBlock = IfBlock->splitBasicBlock(InsertPt, "cond.store");
+ LoopVectorBody.push_back(CondBlock);
+ VectorLp->addBasicBlockToLoop(CondBlock, *LI);
+ // Update Builder with newly created basic block.
+ Builder.SetInsertPoint(InsertPt);
+ }
- // Check if we can replace this instruction with any of the
- // visited instructions.
- if (Instruction *V = CSEMap.lookup(In)) {
- In->replaceAllUsesWith(V);
- In->eraseFromParent();
- continue;
+ Instruction *Cloned = Instr->clone();
+ if (!IsVoidRetTy)
+ Cloned->setName(Instr->getName() + ".cloned");
+ // Replace the operands of the cloned instructions with extracted scalars.
+ for (unsigned op = 0, e = Instr->getNumOperands(); op != e; ++op) {
+ Value *Op = Params[op][Part];
+ // Param is a vector. Need to extract the right lane.
+ if (Op->getType()->isVectorTy())
+ Op = Builder.CreateExtractElement(Op, Builder.getInt32(Width));
+ Cloned->setOperand(op, Op);
}
- // Ignore instructions in conditional blocks. We create "if (pred) a[i] =
- // ...;" blocks for predicated stores. Every second block is a predicated
- // block.
- if (isPredicatedBlock(i))
- continue;
- CSEMap[In] = In;
+ // Place the cloned scalar in the new loop.
+ Builder.Insert(Cloned);
+
+ // If the original scalar returns a value we need to place it in a vector
+ // so that future users will be able to use it.
+ if (!IsVoidRetTy)
+ VecResults[Part] = Builder.CreateInsertElement(VecResults[Part], Cloned,
+ Builder.getInt32(Width));
+ // End if-block.
+ if (IfPredicateStore) {
+ BasicBlock *NewIfBlock = CondBlock->splitBasicBlock(InsertPt, "else");
+ LoopVectorBody.push_back(NewIfBlock);
+ VectorLp->addBasicBlockToLoop(NewIfBlock, *LI);
+ Builder.SetInsertPoint(InsertPt);
+ ReplaceInstWithInst(IfBlock->getTerminator(),
+ BranchInst::Create(CondBlock, NewIfBlock, Cmp));
+ IfBlock = NewIfBlock;
+ }
}
}
}
-/// \brief Adds a 'fast' flag to floating point operations.
-static Value *addFastMathFlag(Value *V) {
- if (isa<FPMathOperator>(V)){
- FastMathFlags Flags;
- Flags.setUnsafeAlgebra();
- cast<Instruction>(V)->setFastMathFlags(Flags);
- }
- return V;
+static Instruction *getFirstInst(Instruction *FirstInst, Value *V,
+ Instruction *Loc) {
+ if (FirstInst)
+ return FirstInst;
+ if (Instruction *I = dyn_cast<Instruction>(V))
+ return I->getParent() == Loc->getParent() ? I : nullptr;
+ return nullptr;
}
-void InnerLoopVectorizer::vectorizeLoop() {
- //===------------------------------------------------===//
- //
- // Notice: any optimization or new instruction that go
- // into the code below should be also be implemented in
- // the cost-model.
- //
- //===------------------------------------------------===//
- Constant *Zero = Builder.getInt32(0);
+std::pair<Instruction *, Instruction *>
+InnerLoopVectorizer::addStrideCheck(Instruction *Loc) {
+ Instruction *tnullptr = nullptr;
+ if (!Legal->mustCheckStrides())
+ return std::pair<Instruction *, Instruction *>(tnullptr, tnullptr);
- // In order to support reduction variables we need to be able to vectorize
- // Phi nodes. Phi nodes have cycles, so we need to vectorize them in two
- // stages. First, we create a new vector PHI node with no incoming edges.
- // We use this value when we vectorize all of the instructions that use the
- // PHI. Next, after all of the instructions in the block are complete we
- // add the new incoming edges to the PHI. At this point all of the
- // instructions in the basic block are vectorized, so we can use them to
- // construct the PHI.
- PhiVector RdxPHIsToFix;
+ IRBuilder<> ChkBuilder(Loc);
- // Scan the loop in a topological order to ensure that defs are vectorized
- // before users.
- LoopBlocksDFS DFS(OrigLoop);
- DFS.perform(LI);
+ // Emit checks.
+ Value *Check = nullptr;
+ Instruction *FirstInst = nullptr;
+ for (SmallPtrSet<Value *, 8>::iterator SI = Legal->strides_begin(),
+ SE = Legal->strides_end();
+ SI != SE; ++SI) {
+ Value *Ptr = stripIntegerCast(*SI);
+ Value *C = ChkBuilder.CreateICmpNE(Ptr, ConstantInt::get(Ptr->getType(), 1),
+ "stride.chk");
+ // Store the first instruction we create.
+ FirstInst = getFirstInst(FirstInst, C, Loc);
+ if (Check)
+ Check = ChkBuilder.CreateOr(Check, C);
+ else
+ Check = C;
+ }
- // Vectorize all of the blocks in the original loop.
- for (LoopBlocksDFS::RPOIterator bb = DFS.beginRPO(),
- be = DFS.endRPO(); bb != be; ++bb)
- vectorizeBlockInLoop(*bb, &RdxPHIsToFix);
+ // We have to do this trickery because the IRBuilder might fold the check to a
+ // constant expression in which case there is no Instruction anchored in a
+ // the block.
+ LLVMContext &Ctx = Loc->getContext();
+ Instruction *TheCheck =
+ BinaryOperator::CreateAnd(Check, ConstantInt::getTrue(Ctx));
+ ChkBuilder.Insert(TheCheck, "stride.not.one");
+ FirstInst = getFirstInst(FirstInst, TheCheck, Loc);
- // At this point every instruction in the original loop is widened to
- // a vector form. We are almost done. Now, we need to fix the PHI nodes
- // that we vectorized. The PHI nodes are currently empty because we did
- // not want to introduce cycles. Notice that the remaining PHI nodes
- // that we need to fix are reduction variables.
+ return std::make_pair(FirstInst, TheCheck);
+}
- // Create the 'reduced' values for each of the induction vars.
- // The reduced values are the vector values that we scalarize and combine
- // after the loop is finished.
- for (PhiVector::iterator it = RdxPHIsToFix.begin(), e = RdxPHIsToFix.end();
- it != e; ++it) {
- PHINode *RdxPhi = *it;
- assert(RdxPhi && "Unable to recover vectorized PHI");
+void InnerLoopVectorizer::createEmptyLoop() {
+ /*
+ In this function we generate a new loop. The new loop will contain
+ the vectorized instructions while the old loop will continue to run the
+ scalar remainder.
- // Find the reduction variable descriptor.
- assert(Legal->getReductionVars()->count(RdxPhi) &&
- "Unable to find the reduction variable");
- LoopVectorizationLegality::ReductionDescriptor RdxDesc =
- (*Legal->getReductionVars())[RdxPhi];
+ [ ] <-- Back-edge taken count overflow check.
+ / |
+ / v
+ | [ ] <-- vector loop bypass (may consist of multiple blocks).
+ | / |
+ | / v
+ || [ ] <-- vector pre header.
+ || |
+ || v
+ || [ ] \
+ || [ ]_| <-- vector loop.
+ || |
+ | \ v
+ | >[ ] <--- middle-block.
+ | / |
+ | / v
+ -|- >[ ] <--- new preheader.
+ | |
+ | v
+ | [ ] \
+ | [ ]_| <-- old scalar loop to handle remainder.
+ \ |
+ \ v
+ >[ ] <-- exit block.
+ ...
+ */
- setDebugLocFromInst(Builder, RdxDesc.StartValue);
+ BasicBlock *OldBasicBlock = OrigLoop->getHeader();
+ BasicBlock *VectorPH = OrigLoop->getLoopPreheader();
+ BasicBlock *ExitBlock = OrigLoop->getExitBlock();
+ assert(VectorPH && "Invalid loop structure");
+ assert(ExitBlock && "Must have an exit block");
- // We need to generate a reduction vector from the incoming scalar.
- // To do so, we need to generate the 'identity' vector and override
- // one of the elements with the incoming scalar reduction. We need
- // to do it in the vector-loop preheader.
- Builder.SetInsertPoint(LoopBypassBlocks[1]->getTerminator());
+ // Some loops have a single integer induction variable, while other loops
+ // don't. One example is c++ iterators that often have multiple pointer
+ // induction variables. In the code below we also support a case where we
+ // don't have a single induction variable.
+ OldInduction = Legal->getInduction();
+ Type *IdxTy = Legal->getWidestInductionType();
- // This is the vector-clone of the value that leaves the loop.
- VectorParts &VectorExit = getVectorValue(RdxDesc.LoopExitInstr);
- Type *VecTy = VectorExit[0]->getType();
+ // Find the loop boundaries.
+ const SCEV *ExitCount = SE->getBackedgeTakenCount(OrigLoop);
+ assert(ExitCount != SE->getCouldNotCompute() && "Invalid loop count");
- // Find the reduction identity variable. Zero for addition, or, xor,
- // one for multiplication, -1 for And.
- Value *Identity;
- Value *VectorStart;
- if (RdxDesc.Kind == LoopVectorizationLegality::RK_IntegerMinMax ||
- RdxDesc.Kind == LoopVectorizationLegality::RK_FloatMinMax) {
- // MinMax reduction have the start value as their identify.
- if (VF == 1) {
- VectorStart = Identity = RdxDesc.StartValue;
- } else {
- VectorStart = Identity = Builder.CreateVectorSplat(VF,
- RdxDesc.StartValue,
- "minmax.ident");
- }
- } else {
- // Handle other reduction kinds:
- Constant *Iden =
- LoopVectorizationLegality::getReductionIdentity(RdxDesc.Kind,
- VecTy->getScalarType());
- if (VF == 1) {
- Identity = Iden;
- // This vector is the Identity vector where the first element is the
- // incoming scalar reduction.
- VectorStart = RdxDesc.StartValue;
- } else {
- Identity = ConstantVector::getSplat(VF, Iden);
+ // The exit count might have the type of i64 while the phi is i32. This can
+ // happen if we have an induction variable that is sign extended before the
+ // compare. The only way that we get a backedge taken count is that the
+ // induction variable was signed and as such will not overflow. In such a case
+ // truncation is legal.
+ if (ExitCount->getType()->getPrimitiveSizeInBits() >
+ IdxTy->getPrimitiveSizeInBits())
+ ExitCount = SE->getTruncateOrNoop(ExitCount, IdxTy);
- // This vector is the Identity vector where the first element is the
- // incoming scalar reduction.
- VectorStart = Builder.CreateInsertElement(Identity,
- RdxDesc.StartValue, Zero);
- }
- }
+ const SCEV *BackedgeTakeCount = SE->getNoopOrZeroExtend(ExitCount, IdxTy);
+ // Get the total trip count from the count by adding 1.
+ ExitCount = SE->getAddExpr(BackedgeTakeCount,
+ SE->getConstant(BackedgeTakeCount->getType(), 1));
- // Fix the vector-loop phi.
+ const DataLayout &DL = OldBasicBlock->getModule()->getDataLayout();
- // Reductions do not have to start at zero. They can start with
- // any loop invariant values.
- VectorParts &VecRdxPhi = WidenMap.get(RdxPhi);
- BasicBlock *Latch = OrigLoop->getLoopLatch();
- Value *LoopVal = RdxPhi->getIncomingValueForBlock(Latch);
- VectorParts &Val = getVectorValue(LoopVal);
- for (unsigned part = 0; part < UF; ++part) {
- // Make sure to add the reduction stat value only to the
- // first unroll part.
- Value *StartVal = (part == 0) ? VectorStart : Identity;
- cast<PHINode>(VecRdxPhi[part])->addIncoming(StartVal,
- LoopVectorPreHeader);
- cast<PHINode>(VecRdxPhi[part])->addIncoming(Val[part],
- LoopVectorBody.back());
- }
+ // Expand the trip count and place the new instructions in the preheader.
+ // Notice that the pre-header does not change, only the loop body.
+ SCEVExpander Exp(*SE, DL, "induction");
- // Before each round, move the insertion point right between
- // the PHIs and the values we are going to write.
- // This allows us to write both PHINodes and the extractelement
- // instructions.
- Builder.SetInsertPoint(LoopMiddleBlock->getFirstInsertionPt());
+ // We need to test whether the backedge-taken count is uint##_max. Adding one
+ // to it will cause overflow and an incorrect loop trip count in the vector
+ // body. In case of overflow we want to directly jump to the scalar remainder
+ // loop.
+ Value *BackedgeCount =
+ Exp.expandCodeFor(BackedgeTakeCount, BackedgeTakeCount->getType(),
+ VectorPH->getTerminator());
+ if (BackedgeCount->getType()->isPointerTy())
+ BackedgeCount = CastInst::CreatePointerCast(BackedgeCount, IdxTy,
+ "backedge.ptrcnt.to.int",
+ VectorPH->getTerminator());
+ Instruction *CheckBCOverflow =
+ CmpInst::Create(Instruction::ICmp, CmpInst::ICMP_EQ, BackedgeCount,
+ Constant::getAllOnesValue(BackedgeCount->getType()),
+ "backedge.overflow", VectorPH->getTerminator());
- VectorParts RdxParts;
- setDebugLocFromInst(Builder, RdxDesc.LoopExitInstr);
- for (unsigned part = 0; part < UF; ++part) {
- // This PHINode contains the vectorized reduction variable, or
- // the initial value vector, if we bypass the vector loop.
- VectorParts &RdxExitVal = getVectorValue(RdxDesc.LoopExitInstr);
- PHINode *NewPhi = Builder.CreatePHI(VecTy, 2, "rdx.vec.exit.phi");
- Value *StartVal = (part == 0) ? VectorStart : Identity;
- for (unsigned I = 1, E = LoopBypassBlocks.size(); I != E; ++I)
- NewPhi->addIncoming(StartVal, LoopBypassBlocks[I]);
- NewPhi->addIncoming(RdxExitVal[part],
- LoopVectorBody.back());
- RdxParts.push_back(NewPhi);
- }
+ // The loop index does not have to start at Zero. Find the original start
+ // value from the induction PHI node. If we don't have an induction variable
+ // then we know that it starts at zero.
+ Builder.SetInsertPoint(VectorPH->getTerminator());
+ Value *StartIdx = ExtendedIdx =
+ OldInduction
+ ? Builder.CreateZExt(OldInduction->getIncomingValueForBlock(VectorPH),
+ IdxTy)
+ : ConstantInt::get(IdxTy, 0);
- // Reduce all of the unrolled parts into a single vector.
- Value *ReducedPartRdx = RdxParts[0];
- unsigned Op = getReductionBinOp(RdxDesc.Kind);
- setDebugLocFromInst(Builder, ReducedPartRdx);
- for (unsigned part = 1; part < UF; ++part) {
- if (Op != Instruction::ICmp && Op != Instruction::FCmp)
- // Floating point operations had to be 'fast' to enable the reduction.
- ReducedPartRdx = addFastMathFlag(
- Builder.CreateBinOp((Instruction::BinaryOps)Op, RdxParts[part],
- ReducedPartRdx, "bin.rdx"));
- else
- ReducedPartRdx = createMinMaxOp(Builder, RdxDesc.MinMaxKind,
- ReducedPartRdx, RdxParts[part]);
- }
+ // Count holds the overall loop count (N).
+ Value *Count = Exp.expandCodeFor(ExitCount, ExitCount->getType(),
+ VectorPH->getTerminator());
- if (VF > 1) {
- // VF is a power of 2 so we can emit the reduction using log2(VF) shuffles
- // and vector ops, reducing the set of values being computed by half each
- // round.
- assert(isPowerOf2_32(VF) &&
- "Reduction emission only supported for pow2 vectors!");
- Value *TmpVec = ReducedPartRdx;
- SmallVector<Constant*, 32> ShuffleMask(VF, nullptr);
- for (unsigned i = VF; i != 1; i >>= 1) {
- // Move the upper half of the vector to the lower half.
- for (unsigned j = 0; j != i/2; ++j)
- ShuffleMask[j] = Builder.getInt32(i/2 + j);
+ LoopBypassBlocks.push_back(VectorPH);
- // Fill the rest of the mask with undef.
- std::fill(&ShuffleMask[i/2], ShuffleMask.end(),
- UndefValue::get(Builder.getInt32Ty()));
+ // Split the single block loop into the two loop structure described above.
+ BasicBlock *VecBody =
+ VectorPH->splitBasicBlock(VectorPH->getTerminator(), "vector.body");
+ BasicBlock *MiddleBlock =
+ VecBody->splitBasicBlock(VecBody->getTerminator(), "middle.block");
+ BasicBlock *ScalarPH =
+ MiddleBlock->splitBasicBlock(MiddleBlock->getTerminator(), "scalar.ph");
- Value *Shuf =
- Builder.CreateShuffleVector(TmpVec,
- UndefValue::get(TmpVec->getType()),
- ConstantVector::get(ShuffleMask),
- "rdx.shuf");
+ // Create and register the new vector loop.
+ Loop* Lp = new Loop();
+ Loop *ParentLoop = OrigLoop->getParentLoop();
- if (Op != Instruction::ICmp && Op != Instruction::FCmp)
- // Floating point operations had to be 'fast' to enable the reduction.
- TmpVec = addFastMathFlag(Builder.CreateBinOp(
- (Instruction::BinaryOps)Op, TmpVec, Shuf, "bin.rdx"));
- else
- TmpVec = createMinMaxOp(Builder, RdxDesc.MinMaxKind, TmpVec, Shuf);
- }
+ // Insert the new loop into the loop nest and register the new basic blocks
+ // before calling any utilities such as SCEV that require valid LoopInfo.
+ if (ParentLoop) {
+ ParentLoop->addChildLoop(Lp);
+ ParentLoop->addBasicBlockToLoop(ScalarPH, *LI);
+ ParentLoop->addBasicBlockToLoop(MiddleBlock, *LI);
+ } else {
+ LI->addTopLevelLoop(Lp);
+ }
+ Lp->addBasicBlockToLoop(VecBody, *LI);
- // The result is in the first element of the vector.
- ReducedPartRdx = Builder.CreateExtractElement(TmpVec,
- Builder.getInt32(0));
- }
+ // Use this IR builder to create the loop instructions (Phi, Br, Cmp)
+ // inside the loop.
+ Builder.SetInsertPoint(VecBody->getFirstNonPHI());
- // Create a phi node that merges control-flow from the backedge-taken check
- // block and the middle block.
- PHINode *BCBlockPhi = PHINode::Create(RdxPhi->getType(), 2, "bc.merge.rdx",
- LoopScalarPreHeader->getTerminator());
- BCBlockPhi->addIncoming(RdxDesc.StartValue, LoopBypassBlocks[0]);
- BCBlockPhi->addIncoming(ReducedPartRdx, LoopMiddleBlock);
+ // Generate the induction variable.
+ setDebugLocFromInst(Builder, getDebugLocFromInstOrOperands(OldInduction));
+ Induction = Builder.CreatePHI(IdxTy, 2, "index");
+ // The loop step is equal to the vectorization factor (num of SIMD elements)
+ // times the unroll factor (num of SIMD instructions).
+ Constant *Step = ConstantInt::get(IdxTy, VF * UF);
- // Now, we need to fix the users of the reduction variable
- // inside and outside of the scalar remainder loop.
- // We know that the loop is in LCSSA form. We need to update the
- // PHI nodes in the exit blocks.
- for (BasicBlock::iterator LEI = LoopExitBlock->begin(),
- LEE = LoopExitBlock->end(); LEI != LEE; ++LEI) {
- PHINode *LCSSAPhi = dyn_cast<PHINode>(LEI);
- if (!LCSSAPhi) break;
+ // Generate code to check that the loop's trip count that we computed by
+ // adding one to the backedge-taken count will not overflow.
+ BasicBlock *NewVectorPH =
+ VectorPH->splitBasicBlock(VectorPH->getTerminator(), "overflow.checked");
+ if (ParentLoop)
+ ParentLoop->addBasicBlockToLoop(NewVectorPH, *LI);
+ ReplaceInstWithInst(
+ VectorPH->getTerminator(),
+ BranchInst::Create(ScalarPH, NewVectorPH, CheckBCOverflow));
+ VectorPH = NewVectorPH;
- // All PHINodes need to have a single entry edge, or two if
- // we already fixed them.
- assert(LCSSAPhi->getNumIncomingValues() < 3 && "Invalid LCSSA PHI");
+ // This is the IR builder that we use to add all of the logic for bypassing
+ // the new vector loop.
+ IRBuilder<> BypassBuilder(VectorPH->getTerminator());
+ setDebugLocFromInst(BypassBuilder,
+ getDebugLocFromInstOrOperands(OldInduction));
- // We found our reduction value exit-PHI. Update it with the
- // incoming bypass edge.
- if (LCSSAPhi->getIncomingValue(0) == RdxDesc.LoopExitInstr) {
- // Add an edge coming from the bypass.
- LCSSAPhi->addIncoming(ReducedPartRdx, LoopMiddleBlock);
- break;
- }
- }// end of the LCSSA phi scan.
-
- // Fix the scalar loop reduction variable with the incoming reduction sum
- // from the vector body and from the backedge value.
- int IncomingEdgeBlockIdx =
- (RdxPhi)->getBasicBlockIndex(OrigLoop->getLoopLatch());
- assert(IncomingEdgeBlockIdx >= 0 && "Invalid block index");
- // Pick the other block.
- int SelfEdgeBlockIdx = (IncomingEdgeBlockIdx ? 0 : 1);
- (RdxPhi)->setIncomingValue(SelfEdgeBlockIdx, BCBlockPhi);
- (RdxPhi)->setIncomingValue(IncomingEdgeBlockIdx, RdxDesc.LoopExitInstr);
- }// end of for each redux variable.
-
- fixLCSSAPHIs();
-
- // Remove redundant induction instructions.
- cse(LoopVectorBody);
-}
-
-void InnerLoopVectorizer::fixLCSSAPHIs() {
- for (BasicBlock::iterator LEI = LoopExitBlock->begin(),
- LEE = LoopExitBlock->end(); LEI != LEE; ++LEI) {
- PHINode *LCSSAPhi = dyn_cast<PHINode>(LEI);
- if (!LCSSAPhi) break;
- if (LCSSAPhi->getNumIncomingValues() == 1)
- LCSSAPhi->addIncoming(UndefValue::get(LCSSAPhi->getType()),
- LoopMiddleBlock);
+ // We may need to extend the index in case there is a type mismatch.
+ // We know that the count starts at zero and does not overflow.
+ if (Count->getType() != IdxTy) {
+ // The exit count can be of pointer type. Convert it to the correct
+ // integer type.
+ if (ExitCount->getType()->isPointerTy())
+ Count = BypassBuilder.CreatePointerCast(Count, IdxTy, "ptrcnt.to.int");
+ else
+ Count = BypassBuilder.CreateZExtOrTrunc(Count, IdxTy, "cnt.cast");
}
-}
-
-InnerLoopVectorizer::VectorParts
-InnerLoopVectorizer::createEdgeMask(BasicBlock *Src, BasicBlock *Dst) {
- assert(std::find(pred_begin(Dst), pred_end(Dst), Src) != pred_end(Dst) &&
- "Invalid edge");
-
- // Look for cached value.
- std::pair<BasicBlock*, BasicBlock*> Edge(Src, Dst);
- EdgeMaskCache::iterator ECEntryIt = MaskCache.find(Edge);
- if (ECEntryIt != MaskCache.end())
- return ECEntryIt->second;
- VectorParts SrcMask = createBlockInMask(Src);
+ // Add the start index to the loop count to get the new end index.
+ Value *IdxEnd = BypassBuilder.CreateAdd(Count, StartIdx, "end.idx");
- // The terminator has to be a branch inst!
- BranchInst *BI = dyn_cast<BranchInst>(Src->getTerminator());
- assert(BI && "Unexpected terminator found");
+ // Now we need to generate the expression for N - (N % VF), which is
+ // the part that the vectorized body will execute.
+ Value *R = BypassBuilder.CreateURem(Count, Step, "n.mod.vf");
+ Value *CountRoundDown = BypassBuilder.CreateSub(Count, R, "n.vec");
+ Value *IdxEndRoundDown = BypassBuilder.CreateAdd(CountRoundDown, StartIdx,
+ "end.idx.rnd.down");
- if (BI->isConditional()) {
- VectorParts EdgeMask = getVectorValue(BI->getCondition());
+ // Now, compare the new count to zero. If it is zero skip the vector loop and
+ // jump to the scalar loop.
+ Value *Cmp =
+ BypassBuilder.CreateICmpEQ(IdxEndRoundDown, StartIdx, "cmp.zero");
+ NewVectorPH =
+ VectorPH->splitBasicBlock(VectorPH->getTerminator(), "vector.ph");
+ if (ParentLoop)
+ ParentLoop->addBasicBlockToLoop(NewVectorPH, *LI);
+ LoopBypassBlocks.push_back(VectorPH);
+ ReplaceInstWithInst(VectorPH->getTerminator(),
+ BranchInst::Create(MiddleBlock, NewVectorPH, Cmp));
+ VectorPH = NewVectorPH;
- if (BI->getSuccessor(0) != Dst)
- for (unsigned part = 0; part < UF; ++part)
- EdgeMask[part] = Builder.CreateNot(EdgeMask[part]);
+ // Generate the code to check that the strides we assumed to be one are really
+ // one. We want the new basic block to start at the first instruction in a
+ // sequence of instructions that form a check.
+ Instruction *StrideCheck;
+ Instruction *FirstCheckInst;
+ std::tie(FirstCheckInst, StrideCheck) =
+ addStrideCheck(VectorPH->getTerminator());
+ if (StrideCheck) {
+ AddedSafetyChecks = true;
+ // Create a new block containing the stride check.
+ VectorPH->setName("vector.stridecheck");
+ NewVectorPH =
+ VectorPH->splitBasicBlock(VectorPH->getTerminator(), "vector.ph");
+ if (ParentLoop)
+ ParentLoop->addBasicBlockToLoop(NewVectorPH, *LI);
+ LoopBypassBlocks.push_back(VectorPH);
- for (unsigned part = 0; part < UF; ++part)
- EdgeMask[part] = Builder.CreateAnd(EdgeMask[part], SrcMask[part]);
+ // Replace the branch into the memory check block with a conditional branch
+ // for the "few elements case".
+ ReplaceInstWithInst(
+ VectorPH->getTerminator(),
+ BranchInst::Create(MiddleBlock, NewVectorPH, StrideCheck));
- MaskCache[Edge] = EdgeMask;
- return EdgeMask;
+ VectorPH = NewVectorPH;
}
- MaskCache[Edge] = SrcMask;
- return SrcMask;
-}
+ // Generate the code that checks in runtime if arrays overlap. We put the
+ // checks into a separate block to make the more common case of few elements
+ // faster.
+ Instruction *MemRuntimeCheck;
+ std::tie(FirstCheckInst, MemRuntimeCheck) =
+ Legal->getLAI()->addRuntimeChecks(VectorPH->getTerminator());
+ if (MemRuntimeCheck) {
+ AddedSafetyChecks = true;
+ // Create a new block containing the memory check.
+ VectorPH->setName("vector.memcheck");
+ NewVectorPH =
+ VectorPH->splitBasicBlock(VectorPH->getTerminator(), "vector.ph");
+ if (ParentLoop)
+ ParentLoop->addBasicBlockToLoop(NewVectorPH, *LI);
+ LoopBypassBlocks.push_back(VectorPH);
-InnerLoopVectorizer::VectorParts
-InnerLoopVectorizer::createBlockInMask(BasicBlock *BB) {
- assert(OrigLoop->contains(BB) && "Block is not a part of a loop");
+ // Replace the branch into the memory check block with a conditional branch
+ // for the "few elements case".
+ ReplaceInstWithInst(
+ VectorPH->getTerminator(),
+ BranchInst::Create(MiddleBlock, NewVectorPH, MemRuntimeCheck));
- // Loop incoming mask is all-one.
- if (OrigLoop->getHeader() == BB) {
- Value *C = ConstantInt::get(IntegerType::getInt1Ty(BB->getContext()), 1);
- return getVectorValue(C);
+ VectorPH = NewVectorPH;
}
- // This is the block mask. We OR all incoming edges, and with zero.
- Value *Zero = ConstantInt::get(IntegerType::getInt1Ty(BB->getContext()), 0);
- VectorParts BlockMask = getVectorValue(Zero);
+ // We are going to resume the execution of the scalar loop.
+ // Go over all of the induction variables that we found and fix the
+ // PHIs that are left in the scalar version of the loop.
+ // The starting values of PHI nodes depend on the counter of the last
+ // iteration in the vectorized loop.
+ // If we come from a bypass edge then we need to start from the original
+ // start value.
- // For each pred:
- for (pred_iterator it = pred_begin(BB), e = pred_end(BB); it != e; ++it) {
- VectorParts EM = createEdgeMask(*it, BB);
- for (unsigned part = 0; part < UF; ++part)
- BlockMask[part] = Builder.CreateOr(BlockMask[part], EM[part]);
- }
+ // This variable saves the new starting index for the scalar loop.
+ PHINode *ResumeIndex = nullptr;
+ LoopVectorizationLegality::InductionList::iterator I, E;
+ LoopVectorizationLegality::InductionList *List = Legal->getInductionVars();
+ // Set builder to point to last bypass block.
+ BypassBuilder.SetInsertPoint(LoopBypassBlocks.back()->getTerminator());
+ for (I = List->begin(), E = List->end(); I != E; ++I) {
+ PHINode *OrigPhi = I->first;
+ LoopVectorizationLegality::InductionInfo II = I->second;
- return BlockMask;
-}
+ Type *ResumeValTy = (OrigPhi == OldInduction) ? IdxTy : OrigPhi->getType();
+ PHINode *ResumeVal = PHINode::Create(ResumeValTy, 2, "resume.val",
+ MiddleBlock->getTerminator());
+ // We might have extended the type of the induction variable but we need a
+ // truncated version for the scalar loop.
+ PHINode *TruncResumeVal = (OrigPhi == OldInduction) ?
+ PHINode::Create(OrigPhi->getType(), 2, "trunc.resume.val",
+ MiddleBlock->getTerminator()) : nullptr;
-void InnerLoopVectorizer::widenPHIInstruction(Instruction *PN,
- InnerLoopVectorizer::VectorParts &Entry,
- unsigned UF, unsigned VF, PhiVector *PV) {
- PHINode* P = cast<PHINode>(PN);
- // Handle reduction variables:
- if (Legal->getReductionVars()->count(P)) {
- for (unsigned part = 0; part < UF; ++part) {
- // This is phase one of vectorizing PHIs.
- Type *VecTy = (VF == 1) ? PN->getType() :
- VectorType::get(PN->getType(), VF);
- Entry[part] = PHINode::Create(VecTy, 2, "vec.phi",
- LoopVectorBody.back()-> getFirstInsertionPt());
+ // Create phi nodes to merge from the backedge-taken check block.
+ PHINode *BCResumeVal = PHINode::Create(ResumeValTy, 3, "bc.resume.val",
+ ScalarPH->getTerminator());
+ BCResumeVal->addIncoming(ResumeVal, MiddleBlock);
+
+ PHINode *BCTruncResumeVal = nullptr;
+ if (OrigPhi == OldInduction) {
+ BCTruncResumeVal =
+ PHINode::Create(OrigPhi->getType(), 2, "bc.trunc.resume.val",
+ ScalarPH->getTerminator());
+ BCTruncResumeVal->addIncoming(TruncResumeVal, MiddleBlock);
}
- PV->push_back(P);
- return;
- }
- setDebugLocFromInst(Builder, P);
- // Check for PHI nodes that are lowered to vector selects.
- if (P->getParent() != OrigLoop->getHeader()) {
- // We know that all PHIs in non-header blocks are converted into
- // selects, so we don't have to worry about the insertion order and we
- // can just use the builder.
- // At this point we generate the predication tree. There may be
- // duplications since this is a simple recursive scan, but future
- // optimizations will clean it up.
+ Value *EndValue = nullptr;
+ switch (II.IK) {
+ case LoopVectorizationLegality::IK_NoInduction:
+ llvm_unreachable("Unknown induction");
+ case LoopVectorizationLegality::IK_IntInduction: {
+ // Handle the integer induction counter.
+ assert(OrigPhi->getType()->isIntegerTy() && "Invalid type");
- unsigned NumIncoming = P->getNumIncomingValues();
+ // We have the canonical induction variable.
+ if (OrigPhi == OldInduction) {
+ // Create a truncated version of the resume value for the scalar loop,
+ // we might have promoted the type to a larger width.
+ EndValue =
+ BypassBuilder.CreateTrunc(IdxEndRoundDown, OrigPhi->getType());
+ // The new PHI merges the original incoming value, in case of a bypass,
+ // or the value at the end of the vectorized loop.
+ for (unsigned I = 1, E = LoopBypassBlocks.size(); I != E; ++I)
+ TruncResumeVal->addIncoming(II.StartValue, LoopBypassBlocks[I]);
+ TruncResumeVal->addIncoming(EndValue, VecBody);
- // Generate a sequence of selects of the form:
- // SELECT(Mask3, In3,
- // SELECT(Mask2, In2,
- // ( ...)))
- for (unsigned In = 0; In < NumIncoming; In++) {
- VectorParts Cond = createEdgeMask(P->getIncomingBlock(In),
- P->getParent());
- VectorParts &In0 = getVectorValue(P->getIncomingValue(In));
+ BCTruncResumeVal->addIncoming(II.StartValue, LoopBypassBlocks[0]);
- for (unsigned part = 0; part < UF; ++part) {
- // We might have single edge PHIs (blocks) - use an identity
- // 'select' for the first PHI operand.
- if (In == 0)
- Entry[part] = Builder.CreateSelect(Cond[part], In0[part],
- In0[part]);
- else
- // Select between the current value and the previous incoming edge
- // based on the incoming mask.
- Entry[part] = Builder.CreateSelect(Cond[part], In0[part],
- Entry[part], "predphi");
+ // We know what the end value is.
+ EndValue = IdxEndRoundDown;
+ // We also know which PHI node holds it.
+ ResumeIndex = ResumeVal;
+ break;
}
+
+ // Not the canonical induction variable - add the vector loop count to the
+ // start value.
+ Value *CRD = BypassBuilder.CreateSExtOrTrunc(CountRoundDown,
+ II.StartValue->getType(),
+ "cast.crd");
+ EndValue = II.transform(BypassBuilder, CRD);
+ EndValue->setName("ind.end");
+ break;
}
- return;
- }
+ case LoopVectorizationLegality::IK_PtrInduction: {
+ Value *CRD = BypassBuilder.CreateSExtOrTrunc(CountRoundDown,
+ II.StepValue->getType(),
+ "cast.crd");
+ EndValue = II.transform(BypassBuilder, CRD);
+ EndValue->setName("ptr.ind.end");
+ break;
+ }
+ }// end of case
- // This PHINode must be an induction variable.
- // Make sure that we know about it.
- assert(Legal->getInductionVars()->count(P) &&
- "Not an induction variable");
+ // The new PHI merges the original incoming value, in case of a bypass,
+ // or the value at the end of the vectorized loop.
+ for (unsigned I = 1, E = LoopBypassBlocks.size(); I != E; ++I) {
+ if (OrigPhi == OldInduction)
+ ResumeVal->addIncoming(StartIdx, LoopBypassBlocks[I]);
+ else
+ ResumeVal->addIncoming(II.StartValue, LoopBypassBlocks[I]);
+ }
+ ResumeVal->addIncoming(EndValue, VecBody);
- LoopVectorizationLegality::InductionInfo II =
- Legal->getInductionVars()->lookup(P);
+ // Fix the scalar body counter (PHI node).
+ unsigned BlockIdx = OrigPhi->getBasicBlockIndex(ScalarPH);
- // FIXME: The newly created binary instructions should contain nsw/nuw flags,
- // which can be found from the original scalar operations.
- switch (II.IK) {
- case LoopVectorizationLegality::IK_NoInduction:
- llvm_unreachable("Unknown induction");
- case LoopVectorizationLegality::IK_IntInduction: {
- assert(P->getType() == II.StartValue->getType() && "Types must match");
- Type *PhiTy = P->getType();
- Value *Broadcasted;
- if (P == OldInduction) {
- // Handle the canonical induction variable. We might have had to
- // extend the type.
- Broadcasted = Builder.CreateTrunc(Induction, PhiTy);
- } else {
- // Handle other induction variables that are now based on the
- // canonical one.
- Value *NormalizedIdx = Builder.CreateSub(Induction, ExtendedIdx,
- "normalized.idx");
- NormalizedIdx = Builder.CreateSExtOrTrunc(NormalizedIdx, PhiTy);
- Broadcasted = II.transform(Builder, NormalizedIdx);
- Broadcasted->setName("offset.idx");
- }
- Broadcasted = getBroadcastInstrs(Broadcasted);
- // After broadcasting the induction variable we need to make the vector
- // consecutive by adding 0, 1, 2, etc.
- for (unsigned part = 0; part < UF; ++part)
- Entry[part] = getStepVector(Broadcasted, VF * part, II.StepValue);
- return;
- }
- case LoopVectorizationLegality::IK_PtrInduction:
- // Handle the pointer induction variable case.
- assert(P->getType()->isPointerTy() && "Unexpected type.");
- // This is the normalized GEP that starts counting at zero.
- Value *NormalizedIdx =
- Builder.CreateSub(Induction, ExtendedIdx, "normalized.idx");
- // This is the vector of results. Notice that we don't generate
- // vector geps because scalar geps result in better code.
- for (unsigned part = 0; part < UF; ++part) {
- if (VF == 1) {
- int EltIndex = part;
- Constant *Idx = ConstantInt::get(Induction->getType(), EltIndex);
- Value *GlobalIdx = Builder.CreateAdd(NormalizedIdx, Idx);
- Value *SclrGep = II.transform(Builder, GlobalIdx);
- SclrGep->setName("next.gep");
- Entry[part] = SclrGep;
- continue;
- }
-
- Value *VecVal = UndefValue::get(VectorType::get(P->getType(), VF));
- for (unsigned int i = 0; i < VF; ++i) {
- int EltIndex = i + part * VF;
- Constant *Idx = ConstantInt::get(Induction->getType(), EltIndex);
- Value *GlobalIdx = Builder.CreateAdd(NormalizedIdx, Idx);
- Value *SclrGep = II.transform(Builder, GlobalIdx);
- SclrGep->setName("next.gep");
- VecVal = Builder.CreateInsertElement(VecVal, SclrGep,
- Builder.getInt32(i),
- "insert.gep");
- }
- Entry[part] = VecVal;
- }
- return;
- }
-}
-
-void InnerLoopVectorizer::vectorizeBlockInLoop(BasicBlock *BB, PhiVector *PV) {
- // For each instruction in the old loop.
- for (BasicBlock::iterator it = BB->begin(), e = BB->end(); it != e; ++it) {
- VectorParts &Entry = WidenMap.get(it);
- switch (it->getOpcode()) {
- case Instruction::Br:
- // Nothing to do for PHIs and BR, since we already took care of the
- // loop control flow instructions.
- continue;
- case Instruction::PHI: {
- // Vectorize PHINodes.
- widenPHIInstruction(it, Entry, UF, VF, PV);
- continue;
- }// End of PHI.
-
- case Instruction::Add:
- case Instruction::FAdd:
- case Instruction::Sub:
- case Instruction::FSub:
- case Instruction::Mul:
- case Instruction::FMul:
- case Instruction::UDiv:
- case Instruction::SDiv:
- case Instruction::FDiv:
- case Instruction::URem:
- case Instruction::SRem:
- case Instruction::FRem:
- case Instruction::Shl:
- case Instruction::LShr:
- case Instruction::AShr:
- case Instruction::And:
- case Instruction::Or:
- case Instruction::Xor: {
- // Just widen binops.
- BinaryOperator *BinOp = dyn_cast<BinaryOperator>(it);
- setDebugLocFromInst(Builder, BinOp);
- VectorParts &A = getVectorValue(it->getOperand(0));
- VectorParts &B = getVectorValue(it->getOperand(1));
-
- // Use this vector value for all users of the original instruction.
- for (unsigned Part = 0; Part < UF; ++Part) {
- Value *V = Builder.CreateBinOp(BinOp->getOpcode(), A[Part], B[Part]);
-
- if (BinaryOperator *VecOp = dyn_cast<BinaryOperator>(V))
- VecOp->copyIRFlags(BinOp);
-
- Entry[Part] = V;
- }
-
- propagateMetadata(Entry, it);
- break;
- }
- case Instruction::Select: {
- // Widen selects.
- // If the selector is loop invariant we can create a select
- // instruction with a scalar condition. Otherwise, use vector-select.
- bool InvariantCond = SE->isLoopInvariant(SE->getSCEV(it->getOperand(0)),
- OrigLoop);
- setDebugLocFromInst(Builder, it);
-
- // The condition can be loop invariant but still defined inside the
- // loop. This means that we can't just use the original 'cond' value.
- // We have to take the 'vectorized' value and pick the first lane.
- // Instcombine will make this a no-op.
- VectorParts &Cond = getVectorValue(it->getOperand(0));
- VectorParts &Op0 = getVectorValue(it->getOperand(1));
- VectorParts &Op1 = getVectorValue(it->getOperand(2));
-
- Value *ScalarCond = (VF == 1) ? Cond[0] :
- Builder.CreateExtractElement(Cond[0], Builder.getInt32(0));
-
- for (unsigned Part = 0; Part < UF; ++Part) {
- Entry[Part] = Builder.CreateSelect(
- InvariantCond ? ScalarCond : Cond[Part],
- Op0[Part],
- Op1[Part]);
- }
-
- propagateMetadata(Entry, it);
- break;
- }
-
- case Instruction::ICmp:
- case Instruction::FCmp: {
- // Widen compares. Generate vector compares.
- bool FCmp = (it->getOpcode() == Instruction::FCmp);
- CmpInst *Cmp = dyn_cast<CmpInst>(it);
- setDebugLocFromInst(Builder, it);
- VectorParts &A = getVectorValue(it->getOperand(0));
- VectorParts &B = getVectorValue(it->getOperand(1));
- for (unsigned Part = 0; Part < UF; ++Part) {
- Value *C = nullptr;
- if (FCmp)
- C = Builder.CreateFCmp(Cmp->getPredicate(), A[Part], B[Part]);
- else
- C = Builder.CreateICmp(Cmp->getPredicate(), A[Part], B[Part]);
- Entry[Part] = C;
- }
-
- propagateMetadata(Entry, it);
- break;
- }
-
- case Instruction::Store:
- case Instruction::Load:
- vectorizeMemoryInstruction(it);
- break;
- case Instruction::ZExt:
- case Instruction::SExt:
- case Instruction::FPToUI:
- case Instruction::FPToSI:
- case Instruction::FPExt:
- case Instruction::PtrToInt:
- case Instruction::IntToPtr:
- case Instruction::SIToFP:
- case Instruction::UIToFP:
- case Instruction::Trunc:
- case Instruction::FPTrunc:
- case Instruction::BitCast: {
- CastInst *CI = dyn_cast<CastInst>(it);
- setDebugLocFromInst(Builder, it);
- /// Optimize the special case where the source is the induction
- /// variable. Notice that we can only optimize the 'trunc' case
- /// because: a. FP conversions lose precision, b. sext/zext may wrap,
- /// c. other casts depend on pointer size.
- if (CI->getOperand(0) == OldInduction &&
- it->getOpcode() == Instruction::Trunc) {
- Value *ScalarCast = Builder.CreateCast(CI->getOpcode(), Induction,
- CI->getType());
- Value *Broadcasted = getBroadcastInstrs(ScalarCast);
- LoopVectorizationLegality::InductionInfo II =
- Legal->getInductionVars()->lookup(OldInduction);
- Constant *Step =
- ConstantInt::getSigned(CI->getType(), II.StepValue->getSExtValue());
- for (unsigned Part = 0; Part < UF; ++Part)
- Entry[Part] = getStepVector(Broadcasted, VF * Part, Step);
- propagateMetadata(Entry, it);
- break;
- }
- /// Vectorize casts.
- Type *DestTy = (VF == 1) ? CI->getType() :
- VectorType::get(CI->getType(), VF);
-
- VectorParts &A = getVectorValue(it->getOperand(0));
- for (unsigned Part = 0; Part < UF; ++Part)
- Entry[Part] = Builder.CreateCast(CI->getOpcode(), A[Part], DestTy);
- propagateMetadata(Entry, it);
- break;
- }
-
- case Instruction::Call: {
- // Ignore dbg intrinsics.
- if (isa<DbgInfoIntrinsic>(it))
- break;
- setDebugLocFromInst(Builder, it);
-
- Module *M = BB->getParent()->getParent();
- CallInst *CI = cast<CallInst>(it);
- Intrinsic::ID ID = getIntrinsicIDForCall(CI, TLI);
- assert(ID && "Not an intrinsic call!");
- switch (ID) {
- case Intrinsic::assume:
- case Intrinsic::lifetime_end:
- case Intrinsic::lifetime_start:
- scalarizeInstruction(it);
- break;
- default:
- bool HasScalarOpd = hasVectorInstrinsicScalarOpd(ID, 1);
- for (unsigned Part = 0; Part < UF; ++Part) {
- SmallVector<Value *, 4> Args;
- for (unsigned i = 0, ie = CI->getNumArgOperands(); i != ie; ++i) {
- if (HasScalarOpd && i == 1) {
- Args.push_back(CI->getArgOperand(i));
- continue;
- }
- VectorParts &Arg = getVectorValue(CI->getArgOperand(i));
- Args.push_back(Arg[Part]);
- }
- Type *Tys[] = {CI->getType()};
- if (VF > 1)
- Tys[0] = VectorType::get(CI->getType()->getScalarType(), VF);
-
- Function *F = Intrinsic::getDeclaration(M, ID, Tys);
- Entry[Part] = Builder.CreateCall(F, Args);
- }
-
- propagateMetadata(Entry, it);
- break;
- }
- break;
- }
-
- default:
- // All other instructions are unsupported. Scalarize them.
- scalarizeInstruction(it);
- break;
- }// end of switch.
- }// end of for_each instr.
-}
-
-void InnerLoopVectorizer::updateAnalysis() {
- // Forget the original basic block.
- SE->forgetLoop(OrigLoop);
-
- // Update the dominator tree information.
- assert(DT->properlyDominates(LoopBypassBlocks.front(), LoopExitBlock) &&
- "Entry does not dominate exit.");
-
- for (unsigned I = 1, E = LoopBypassBlocks.size(); I != E; ++I)
- DT->addNewBlock(LoopBypassBlocks[I], LoopBypassBlocks[I-1]);
- DT->addNewBlock(LoopVectorPreHeader, LoopBypassBlocks.back());
-
- // Due to if predication of stores we might create a sequence of "if(pred)
- // a[i] = ...; " blocks.
- for (unsigned i = 0, e = LoopVectorBody.size(); i != e; ++i) {
- if (i == 0)
- DT->addNewBlock(LoopVectorBody[0], LoopVectorPreHeader);
- else if (isPredicatedBlock(i)) {
- DT->addNewBlock(LoopVectorBody[i], LoopVectorBody[i-1]);
+ // The old induction's phi node in the scalar body needs the truncated
+ // value.
+ if (OrigPhi == OldInduction) {
+ BCResumeVal->addIncoming(StartIdx, LoopBypassBlocks[0]);
+ OrigPhi->setIncomingValue(BlockIdx, BCTruncResumeVal);
} else {
- DT->addNewBlock(LoopVectorBody[i], LoopVectorBody[i-2]);
- }
- }
-
- DT->addNewBlock(LoopMiddleBlock, LoopBypassBlocks[1]);
- DT->addNewBlock(LoopScalarPreHeader, LoopBypassBlocks[0]);
- DT->changeImmediateDominator(LoopScalarBody, LoopScalarPreHeader);
- DT->changeImmediateDominator(LoopExitBlock, LoopBypassBlocks[0]);
-
- DEBUG(DT->verifyDomTree());
-}
-
-/// \brief Check whether it is safe to if-convert this phi node.
-///
-/// Phi nodes with constant expressions that can trap are not safe to if
-/// convert.
-static bool canIfConvertPHINodes(BasicBlock *BB) {
- for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
- PHINode *Phi = dyn_cast<PHINode>(I);
- if (!Phi)
- return true;
- for (unsigned p = 0, e = Phi->getNumIncomingValues(); p != e; ++p)
- if (Constant *C = dyn_cast<Constant>(Phi->getIncomingValue(p)))
- if (C->canTrap())
- return false;
- }
- return true;
-}
-
-bool LoopVectorizationLegality::canVectorizeWithIfConvert() {
- if (!EnableIfConversion) {
- emitAnalysis(VectorizationReport() << "if-conversion is disabled");
- return false;
- }
-
- assert(TheLoop->getNumBlocks() > 1 && "Single block loops are vectorizable");
-
- // A list of pointers that we can safely read and write to.
- SmallPtrSet<Value *, 8> SafePointes;
-
- // Collect safe addresses.
- for (Loop::block_iterator BI = TheLoop->block_begin(),
- BE = TheLoop->block_end(); BI != BE; ++BI) {
- BasicBlock *BB = *BI;
-
- if (blockNeedsPredication(BB))
- continue;
-
- for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
- if (LoadInst *LI = dyn_cast<LoadInst>(I))
- SafePointes.insert(LI->getPointerOperand());
- else if (StoreInst *SI = dyn_cast<StoreInst>(I))
- SafePointes.insert(SI->getPointerOperand());
- }
- }
-
- // Collect the blocks that need predication.
- BasicBlock *Header = TheLoop->getHeader();
- for (Loop::block_iterator BI = TheLoop->block_begin(),
- BE = TheLoop->block_end(); BI != BE; ++BI) {
- BasicBlock *BB = *BI;
-
- // We don't support switch statements inside loops.
- if (!isa<BranchInst>(BB->getTerminator())) {
- emitAnalysis(VectorizationReport(BB->getTerminator())
- << "loop contains a switch statement");
- return false;
- }
-
- // We must be able to predicate all blocks that need to be predicated.
- if (blockNeedsPredication(BB)) {
- if (!blockCanBePredicated(BB, SafePointes)) {
- emitAnalysis(VectorizationReport(BB->getTerminator())
- << "control flow cannot be substituted for a select");
- return false;
- }
- } else if (BB != Header && !canIfConvertPHINodes(BB)) {
- emitAnalysis(VectorizationReport(BB->getTerminator())
- << "control flow cannot be substituted for a select");
- return false;
+ BCResumeVal->addIncoming(II.StartValue, LoopBypassBlocks[0]);
+ OrigPhi->setIncomingValue(BlockIdx, BCResumeVal);
}
}
- // We can if-convert this loop.
- return true;
-}
-
-bool LoopVectorizationLegality::canVectorize() {
- // We must have a loop in canonical form. Loops with indirectbr in them cannot
- // be canonicalized.
- if (!TheLoop->getLoopPreheader()) {
- emitAnalysis(
- VectorizationReport() <<
- "loop control flow is not understood by vectorizer");
- return false;
- }
-
- // We can only vectorize innermost loops.
- if (!TheLoop->getSubLoopsVector().empty()) {
- emitAnalysis(VectorizationReport() << "loop is not the innermost loop");
- return false;
- }
-
- // We must have a single backedge.
- if (TheLoop->getNumBackEdges() != 1) {
- emitAnalysis(
- VectorizationReport() <<
- "loop control flow is not understood by vectorizer");
- return false;
- }
-
- // We must have a single exiting block.
- if (!TheLoop->getExitingBlock()) {
- emitAnalysis(
- VectorizationReport() <<
- "loop control flow is not understood by vectorizer");
- return false;
- }
-
- // We only handle bottom-tested loops, i.e. loop in which the condition is
- // checked at the end of each iteration. With that we can assume that all
- // instructions in the loop are executed the same number of times.
- if (TheLoop->getExitingBlock() != TheLoop->getLoopLatch()) {
- emitAnalysis(
- VectorizationReport() <<
- "loop control flow is not understood by vectorizer");
- return false;
- }
-
- // We need to have a loop header.
- DEBUG(dbgs() << "LV: Found a loop: " <<
- TheLoop->getHeader()->getName() << '\n');
-
- // Check if we can if-convert non-single-bb loops.
- unsigned NumBlocks = TheLoop->getNumBlocks();
- if (NumBlocks != 1 && !canVectorizeWithIfConvert()) {
- DEBUG(dbgs() << "LV: Can't if-convert the loop.\n");
- return false;
- }
-
- // ScalarEvolution needs to be able to find the exit count.
- const SCEV *ExitCount = SE->getBackedgeTakenCount(TheLoop);
- if (ExitCount == SE->getCouldNotCompute()) {
- emitAnalysis(VectorizationReport() <<
- "could not determine number of loop iterations");
- DEBUG(dbgs() << "LV: SCEV could not compute the loop exit count.\n");
- return false;
- }
-
- // Check if we can vectorize the instructions and CFG in this loop.
- if (!canVectorizeInstrs()) {
- DEBUG(dbgs() << "LV: Can't vectorize the instructions or CFG\n");
- return false;
- }
-
- // Go over each instruction and look at memory deps.
- if (!canVectorizeMemory()) {
- DEBUG(dbgs() << "LV: Can't vectorize due to memory conflicts\n");
- return false;
+ // If we are generating a new induction variable then we also need to
+ // generate the code that calculates the exit value. This value is not
+ // simply the end of the counter because we may skip the vectorized body
+ // in case of a runtime check.
+ if (!OldInduction){
+ assert(!ResumeIndex && "Unexpected resume value found");
+ ResumeIndex = PHINode::Create(IdxTy, 2, "new.indc.resume.val",
+ MiddleBlock->getTerminator());
+ for (unsigned I = 1, E = LoopBypassBlocks.size(); I != E; ++I)
+ ResumeIndex->addIncoming(StartIdx, LoopBypassBlocks[I]);
+ ResumeIndex->addIncoming(IdxEndRoundDown, VecBody);
}
- // Collect all of the variables that remain uniform after vectorization.
- collectLoopUniforms();
+ // Make sure that we found the index where scalar loop needs to continue.
+ assert(ResumeIndex && ResumeIndex->getType()->isIntegerTy() &&
+ "Invalid resume Index");
- DEBUG(dbgs() << "LV: We can vectorize this loop" <<
- (LAA.getRuntimePointerCheck()->Need ? " (with a runtime bound check)" :
- "")
- <<"!\n");
+ // Add a check in the middle block to see if we have completed
+ // all of the iterations in the first vector loop.
+ // If (N - N%VF) == N, then we *don't* need to run the remainder.
+ Value *CmpN = CmpInst::Create(Instruction::ICmp, CmpInst::ICMP_EQ, IdxEnd,
+ ResumeIndex, "cmp.n",
+ MiddleBlock->getTerminator());
+ ReplaceInstWithInst(MiddleBlock->getTerminator(),
+ BranchInst::Create(ExitBlock, ScalarPH, CmpN));
- // Okay! We can vectorize. At this point we don't have any other mem analysis
- // which may limit our maximum vectorization factor, so just return true with
- // no restrictions.
- return true;
-}
+ // Create i+1 and fill the PHINode.
+ Value *NextIdx = Builder.CreateAdd(Induction, Step, "index.next");
+ Induction->addIncoming(StartIdx, VectorPH);
+ Induction->addIncoming(NextIdx, VecBody);
+ // Create the compare.
+ Value *ICmp = Builder.CreateICmpEQ(NextIdx, IdxEndRoundDown);
+ Builder.CreateCondBr(ICmp, MiddleBlock, VecBody);
-static Type *convertPointerToIntegerType(const DataLayout &DL, Type *Ty) {
- if (Ty->isPointerTy())
- return DL.getIntPtrType(Ty);
+ // Now we have two terminators. Remove the old one from the block.
+ VecBody->getTerminator()->eraseFromParent();
- // It is possible that char's or short's overflow when we ask for the loop's
- // trip count, work around this by changing the type size.
- if (Ty->getScalarSizeInBits() < 32)
- return Type::getInt32Ty(Ty->getContext());
+ // Get ready to start creating new instructions into the vectorized body.
+ Builder.SetInsertPoint(VecBody->getFirstInsertionPt());
- return Ty;
-}
+ // Save the state.
+ LoopVectorPreHeader = VectorPH;
+ LoopScalarPreHeader = ScalarPH;
+ LoopMiddleBlock = MiddleBlock;
+ LoopExitBlock = ExitBlock;
+ LoopVectorBody.push_back(VecBody);
+ LoopScalarBody = OldBasicBlock;
-static Type* getWiderType(const DataLayout &DL, Type *Ty0, Type *Ty1) {
- Ty0 = convertPointerToIntegerType(DL, Ty0);
- Ty1 = convertPointerToIntegerType(DL, Ty1);
- if (Ty0->getScalarSizeInBits() > Ty1->getScalarSizeInBits())
- return Ty0;
- return Ty1;
+ LoopVectorizeHints Hints(Lp, true);
+ Hints.setAlreadyVectorized();
}
-/// \brief Check that the instruction has outside loop users and is not an
-/// identified reduction variable.
-static bool hasOutsideLoopUser(const Loop *TheLoop, Instruction *Inst,
- SmallPtrSetImpl<Value *> &Reductions) {
- // Reduction instructions are allowed to have exit users. All other
- // instructions must not have external users.
- if (!Reductions.count(Inst))
- //Check that all of the users of the loop are inside the BB.
- for (User *U : Inst->users()) {
- Instruction *UI = cast<Instruction>(U);
- // This user may be a reduction exit value.
- if (!TheLoop->contains(UI)) {
- DEBUG(dbgs() << "LV: Found an outside user for : " << *UI << '\n');
- return true;
- }
- }
- return false;
+namespace {
+struct CSEDenseMapInfo {
+ static bool canHandle(Instruction *I) {
+ return isa<InsertElementInst>(I) || isa<ExtractElementInst>(I) ||
+ isa<ShuffleVectorInst>(I) || isa<GetElementPtrInst>(I);
+ }
+ static inline Instruction *getEmptyKey() {
+ return DenseMapInfo<Instruction *>::getEmptyKey();
+ }
+ static inline Instruction *getTombstoneKey() {
+ return DenseMapInfo<Instruction *>::getTombstoneKey();
+ }
+ static unsigned getHashValue(Instruction *I) {
+ assert(canHandle(I) && "Unknown instruction!");
+ return hash_combine(I->getOpcode(), hash_combine_range(I->value_op_begin(),
+ I->value_op_end()));
+ }
+ static bool isEqual(Instruction *LHS, Instruction *RHS) {
+ if (LHS == getEmptyKey() || RHS == getEmptyKey() ||
+ LHS == getTombstoneKey() || RHS == getTombstoneKey())
+ return LHS == RHS;
+ return LHS->isIdenticalTo(RHS);
+ }
+};
}
-bool LoopVectorizationLegality::canVectorizeInstrs() {
- BasicBlock *PreHeader = TheLoop->getLoopPreheader();
- BasicBlock *Header = TheLoop->getHeader();
-
- // Look for the attribute signaling the absence of NaNs.
- Function &F = *Header->getParent();
- if (F.hasFnAttribute("no-nans-fp-math"))
- HasFunNoNaNAttr = F.getAttributes().getAttribute(
- AttributeSet::FunctionIndex,
- "no-nans-fp-math").getValueAsString() == "true";
-
- // For each block in the loop.
- for (Loop::block_iterator bb = TheLoop->block_begin(),
- be = TheLoop->block_end(); bb != be; ++bb) {
-
- // Scan the instructions in the block and look for hazards.
- for (BasicBlock::iterator it = (*bb)->begin(), e = (*bb)->end(); it != e;
- ++it) {
-
- if (PHINode *Phi = dyn_cast<PHINode>(it)) {
- Type *PhiTy = Phi->getType();
- // Check that this PHI type is allowed.
- if (!PhiTy->isIntegerTy() &&
- !PhiTy->isFloatingPointTy() &&
- !PhiTy->isPointerTy()) {
- emitAnalysis(VectorizationReport(it)
- << "loop control flow is not understood by vectorizer");
- DEBUG(dbgs() << "LV: Found an non-int non-pointer PHI.\n");
- return false;
- }
-
- // If this PHINode is not in the header block, then we know that we
- // can convert it to select during if-conversion. No need to check if
- // the PHIs in this block are induction or reduction variables.
- if (*bb != Header) {
- // Check that this instruction has no outside users or is an
- // identified reduction value with an outside user.
- if (!hasOutsideLoopUser(TheLoop, it, AllowedExit))
- continue;
- emitAnalysis(VectorizationReport(it) <<
- "value could not be identified as "
- "an induction or reduction variable");
- return false;
- }
-
- // We only allow if-converted PHIs with exactly two incoming values.
- if (Phi->getNumIncomingValues() != 2) {
- emitAnalysis(VectorizationReport(it)
- << "control flow not understood by vectorizer");
- DEBUG(dbgs() << "LV: Found an invalid PHI.\n");
- return false;
- }
-
- // This is the value coming from the preheader.
- Value *StartValue = Phi->getIncomingValueForBlock(PreHeader);
- ConstantInt *StepValue = nullptr;
- // Check if this is an induction variable.
- InductionKind IK = isInductionVariable(Phi, StepValue);
-
- if (IK_NoInduction != IK) {
- // Get the widest type.
- if (!WidestIndTy)
- WidestIndTy = convertPointerToIntegerType(*DL, PhiTy);
- else
- WidestIndTy = getWiderType(*DL, PhiTy, WidestIndTy);
-
- // Int inductions are special because we only allow one IV.
- if (IK == IK_IntInduction && StepValue->isOne()) {
- // Use the phi node with the widest type as induction. Use the last
- // one if there are multiple (no good reason for doing this other
- // than it is expedient).
- if (!Induction || PhiTy == WidestIndTy)
- Induction = Phi;
- }
-
- DEBUG(dbgs() << "LV: Found an induction variable.\n");
- Inductions[Phi] = InductionInfo(StartValue, IK, StepValue);
-
- // Until we explicitly handle the case of an induction variable with
- // an outside loop user we have to give up vectorizing this loop.
- if (hasOutsideLoopUser(TheLoop, it, AllowedExit)) {
- emitAnalysis(VectorizationReport(it) <<
- "use of induction value outside of the "
- "loop is not handled by vectorizer");
- return false;
- }
-
- continue;
- }
-
- if (AddReductionVar(Phi, RK_IntegerAdd)) {
- DEBUG(dbgs() << "LV: Found an ADD reduction PHI."<< *Phi <<"\n");
- continue;
- }
- if (AddReductionVar(Phi, RK_IntegerMult)) {
- DEBUG(dbgs() << "LV: Found a MUL reduction PHI."<< *Phi <<"\n");
- continue;
- }
- if (AddReductionVar(Phi, RK_IntegerOr)) {
- DEBUG(dbgs() << "LV: Found an OR reduction PHI."<< *Phi <<"\n");
- continue;
- }
- if (AddReductionVar(Phi, RK_IntegerAnd)) {
- DEBUG(dbgs() << "LV: Found an AND reduction PHI."<< *Phi <<"\n");
- continue;
- }
- if (AddReductionVar(Phi, RK_IntegerXor)) {
- DEBUG(dbgs() << "LV: Found a XOR reduction PHI."<< *Phi <<"\n");
- continue;
- }
- if (AddReductionVar(Phi, RK_IntegerMinMax)) {
- DEBUG(dbgs() << "LV: Found a MINMAX reduction PHI."<< *Phi <<"\n");
- continue;
- }
- if (AddReductionVar(Phi, RK_FloatMult)) {
- DEBUG(dbgs() << "LV: Found an FMult reduction PHI."<< *Phi <<"\n");
- continue;
- }
- if (AddReductionVar(Phi, RK_FloatAdd)) {
- DEBUG(dbgs() << "LV: Found an FAdd reduction PHI."<< *Phi <<"\n");
- continue;
- }
- if (AddReductionVar(Phi, RK_FloatMinMax)) {
- DEBUG(dbgs() << "LV: Found an float MINMAX reduction PHI."<< *Phi <<
- "\n");
- continue;
- }
-
- emitAnalysis(VectorizationReport(it) <<
- "value that could not be identified as "
- "reduction is used outside the loop");
- DEBUG(dbgs() << "LV: Found an unidentified PHI."<< *Phi <<"\n");
- return false;
- }// end of PHI handling
-
- // We still don't handle functions. However, we can ignore dbg intrinsic
- // calls and we do handle certain intrinsic and libm functions.
- CallInst *CI = dyn_cast<CallInst>(it);
- if (CI && !getIntrinsicIDForCall(CI, TLI) && !isa<DbgInfoIntrinsic>(CI)) {
- emitAnalysis(VectorizationReport(it) <<
- "call instruction cannot be vectorized");
- DEBUG(dbgs() << "LV: Found a call site.\n");
- return false;
- }
-
- // Intrinsics such as powi,cttz and ctlz are legal to vectorize if the
- // second argument is the same (i.e. loop invariant)
- if (CI &&
- hasVectorInstrinsicScalarOpd(getIntrinsicIDForCall(CI, TLI), 1)) {
- if (!SE->isLoopInvariant(SE->getSCEV(CI->getOperand(1)), TheLoop)) {
- emitAnalysis(VectorizationReport(it)
- << "intrinsic instruction cannot be vectorized");
- DEBUG(dbgs() << "LV: Found unvectorizable intrinsic " << *CI << "\n");
- return false;
- }
- }
-
- // Check that the instruction return type is vectorizable.
- // Also, we can't vectorize extractelement instructions.
- if ((!VectorType::isValidElementType(it->getType()) &&
- !it->getType()->isVoidTy()) || isa<ExtractElementInst>(it)) {
- emitAnalysis(VectorizationReport(it)
- << "instruction return type cannot be vectorized");
- DEBUG(dbgs() << "LV: Found unvectorizable type.\n");
- return false;
- }
-
- // Check that the stored type is vectorizable.
- if (StoreInst *ST = dyn_cast<StoreInst>(it)) {
- Type *T = ST->getValueOperand()->getType();
- if (!VectorType::isValidElementType(T)) {
- emitAnalysis(VectorizationReport(ST) <<
- "store instruction cannot be vectorized");
- return false;
- }
- if (EnableMemAccessVersioning)
- collectStridedAccess(ST);
- }
-
- if (EnableMemAccessVersioning)
- if (LoadInst *LI = dyn_cast<LoadInst>(it))
- collectStridedAccess(LI);
+/// \brief Check whether this block is a predicated block.
+/// Due to if predication of stores we might create a sequence of "if(pred) a[i]
+/// = ...; " blocks. We start with one vectorized basic block. For every
+/// conditional block we split this vectorized block. Therefore, every second
+/// block will be a predicated one.
+static bool isPredicatedBlock(unsigned BlockNum) {
+ return BlockNum % 2;
+}
- // Reduction instructions are allowed to have exit users.
- // All other instructions must not have external users.
- if (hasOutsideLoopUser(TheLoop, it, AllowedExit)) {
- emitAnalysis(VectorizationReport(it) <<
- "value cannot be used outside the loop");
- return false;
- }
+///\brief Perform cse of induction variable instructions.
+static void cse(SmallVector<BasicBlock *, 4> &BBs) {
+ // Perform simple cse.
+ SmallDenseMap<Instruction *, Instruction *, 4, CSEDenseMapInfo> CSEMap;
+ for (unsigned i = 0, e = BBs.size(); i != e; ++i) {
+ BasicBlock *BB = BBs[i];
+ for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E;) {
+ Instruction *In = I++;
- } // next instr.
+ if (!CSEDenseMapInfo::canHandle(In))
+ continue;
- }
+ // Check if we can replace this instruction with any of the
+ // visited instructions.
+ if (Instruction *V = CSEMap.lookup(In)) {
+ In->replaceAllUsesWith(V);
+ In->eraseFromParent();
+ continue;
+ }
+ // Ignore instructions in conditional blocks. We create "if (pred) a[i] =
+ // ...;" blocks for predicated stores. Every second block is a predicated
+ // block.
+ if (isPredicatedBlock(i))
+ continue;
- if (!Induction) {
- DEBUG(dbgs() << "LV: Did not find one integer induction var.\n");
- if (Inductions.empty()) {
- emitAnalysis(VectorizationReport()
- << "loop induction variable could not be identified");
- return false;
+ CSEMap[In] = In;
}
}
+}
- return true;
+/// \brief Adds a 'fast' flag to floating point operations.
+static Value *addFastMathFlag(Value *V) {
+ if (isa<FPMathOperator>(V)){
+ FastMathFlags Flags;
+ Flags.setUnsafeAlgebra();
+ cast<Instruction>(V)->setFastMathFlags(Flags);
+ }
+ return V;
}
-///\brief Remove GEPs whose indices but the last one are loop invariant and
-/// return the induction operand of the gep pointer.
-static Value *stripGetElementPtr(Value *Ptr, ScalarEvolution *SE,
- const DataLayout *DL, Loop *Lp) {
- GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Ptr);
- if (!GEP)
- return Ptr;
+/// Estimate the overhead of scalarizing a value. Insert and Extract are set if
+/// the result needs to be inserted and/or extracted from vectors.
+static unsigned getScalarizationOverhead(Type *Ty, bool Insert, bool Extract,
+ const TargetTransformInfo &TTI) {
+ if (Ty->isVoidTy())
+ return 0;
- unsigned InductionOperand = getGEPInductionOperand(DL, GEP);
+ assert(Ty->isVectorTy() && "Can only scalarize vectors");
+ unsigned Cost = 0;
- // Check that all of the gep indices are uniform except for our induction
- // operand.
- for (unsigned i = 0, e = GEP->getNumOperands(); i != e; ++i)
- if (i != InductionOperand &&
- !SE->isLoopInvariant(SE->getSCEV(GEP->getOperand(i)), Lp))
- return Ptr;
- return GEP->getOperand(InductionOperand);
+ for (int i = 0, e = Ty->getVectorNumElements(); i < e; ++i) {
+ if (Insert)
+ Cost += TTI.getVectorInstrCost(Instruction::InsertElement, Ty, i);
+ if (Extract)
+ Cost += TTI.getVectorInstrCost(Instruction::ExtractElement, Ty, i);
+ }
+
+ return Cost;
}
-///\brief Look for a cast use of the passed value.
-static Value *getUniqueCastUse(Value *Ptr, Loop *Lp, Type *Ty) {
- Value *UniqueCast = nullptr;
- for (User *U : Ptr->users()) {
- CastInst *CI = dyn_cast<CastInst>(U);
- if (CI && CI->getType() == Ty) {
- if (!UniqueCast)
- UniqueCast = CI;
- else
- return nullptr;
- }
+// Estimate cost of a call instruction CI if it were vectorized with factor VF.
+// Return the cost of the instruction, including scalarization overhead if it's
+// needed. The flag NeedToScalarize shows if the call needs to be scalarized -
+// i.e. either vector version isn't available, or is too expensive.
+static unsigned getVectorCallCost(CallInst *CI, unsigned VF,
+ const TargetTransformInfo &TTI,
+ const TargetLibraryInfo *TLI,
+ bool &NeedToScalarize) {
+ Function *F = CI->getCalledFunction();
+ StringRef FnName = CI->getCalledFunction()->getName();
+ Type *ScalarRetTy = CI->getType();
+ SmallVector<Type *, 4> Tys, ScalarTys;
+ for (auto &ArgOp : CI->arg_operands())
+ ScalarTys.push_back(ArgOp->getType());
+
+ // Estimate cost of scalarized vector call. The source operands are assumed
+ // to be vectors, so we need to extract individual elements from there,
+ // execute VF scalar calls, and then gather the result into the vector return
+ // value.
+ unsigned ScalarCallCost = TTI.getCallInstrCost(F, ScalarRetTy, ScalarTys);
+ if (VF == 1)
+ return ScalarCallCost;
+
+ // Compute corresponding vector type for return value and arguments.
+ Type *RetTy = ToVectorTy(ScalarRetTy, VF);
+ for (unsigned i = 0, ie = ScalarTys.size(); i != ie; ++i)
+ Tys.push_back(ToVectorTy(ScalarTys[i], VF));
+
+ // Compute costs of unpacking argument values for the scalar calls and
+ // packing the return values to a vector.
+ unsigned ScalarizationCost =
+ getScalarizationOverhead(RetTy, true, false, TTI);
+ for (unsigned i = 0, ie = Tys.size(); i != ie; ++i)
+ ScalarizationCost += getScalarizationOverhead(Tys[i], false, true, TTI);
+
+ unsigned Cost = ScalarCallCost * VF + ScalarizationCost;
+
+ // If we can't emit a vector call for this function, then the currently found
+ // cost is the cost we need to return.
+ NeedToScalarize = true;
+ if (!TLI || !TLI->isFunctionVectorizable(FnName, VF) || CI->isNoBuiltin())
+ return Cost;
+
+ // If the corresponding vector cost is cheaper, return its cost.
+ unsigned VectorCallCost = TTI.getCallInstrCost(nullptr, RetTy, Tys);
+ if (VectorCallCost < Cost) {
+ NeedToScalarize = false;
+ return VectorCallCost;
}
- return UniqueCast;
+ return Cost;
}
-///\brief Get the stride of a pointer access in a loop.
-/// Looks for symbolic strides "a[i*stride]". Returns the symbolic stride as a
-/// pointer to the Value, or null otherwise.
-static Value *getStrideFromPointer(Value *Ptr, ScalarEvolution *SE,
- const DataLayout *DL, Loop *Lp) {
- const PointerType *PtrTy = dyn_cast<PointerType>(Ptr->getType());
- if (!PtrTy || PtrTy->isAggregateType())
- return nullptr;
+// Estimate cost of an intrinsic call instruction CI if it were vectorized with
+// factor VF. Return the cost of the instruction, including scalarization
+// overhead if it's needed.
+static unsigned getVectorIntrinsicCost(CallInst *CI, unsigned VF,
+ const TargetTransformInfo &TTI,
+ const TargetLibraryInfo *TLI) {
+ Intrinsic::ID ID = getIntrinsicIDForCall(CI, TLI);
+ assert(ID && "Expected intrinsic call!");
+
+ Type *RetTy = ToVectorTy(CI->getType(), VF);
+ SmallVector<Type *, 4> Tys;
+ for (unsigned i = 0, ie = CI->getNumArgOperands(); i != ie; ++i)
+ Tys.push_back(ToVectorTy(CI->getArgOperand(i)->getType(), VF));
+
+ return TTI.getIntrinsicInstrCost(ID, RetTy, Tys);
+}
+
+void InnerLoopVectorizer::vectorizeLoop() {
+ //===------------------------------------------------===//
+ //
+ // Notice: any optimization or new instruction that go
+ // into the code below should be also be implemented in
+ // the cost-model.
+ //
+ //===------------------------------------------------===//
+ Constant *Zero = Builder.getInt32(0);
- // Try to remove a gep instruction to make the pointer (actually index at this
- // point) easier analyzable. If OrigPtr is equal to Ptr we are analzying the
- // pointer, otherwise, we are analyzing the index.
- Value *OrigPtr = Ptr;
+ // In order to support reduction variables we need to be able to vectorize
+ // Phi nodes. Phi nodes have cycles, so we need to vectorize them in two
+ // stages. First, we create a new vector PHI node with no incoming edges.
+ // We use this value when we vectorize all of the instructions that use the
+ // PHI. Next, after all of the instructions in the block are complete we
+ // add the new incoming edges to the PHI. At this point all of the
+ // instructions in the basic block are vectorized, so we can use them to
+ // construct the PHI.
+ PhiVector RdxPHIsToFix;
- // The size of the pointer access.
- int64_t PtrAccessSize = 1;
+ // Scan the loop in a topological order to ensure that defs are vectorized
+ // before users.
+ LoopBlocksDFS DFS(OrigLoop);
+ DFS.perform(LI);
- Ptr = stripGetElementPtr(Ptr, SE, DL, Lp);
- const SCEV *V = SE->getSCEV(Ptr);
+ // Vectorize all of the blocks in the original loop.
+ for (LoopBlocksDFS::RPOIterator bb = DFS.beginRPO(),
+ be = DFS.endRPO(); bb != be; ++bb)
+ vectorizeBlockInLoop(*bb, &RdxPHIsToFix);
- if (Ptr != OrigPtr)
- // Strip off casts.
- while (const SCEVCastExpr *C = dyn_cast<SCEVCastExpr>(V))
- V = C->getOperand();
+ // At this point every instruction in the original loop is widened to
+ // a vector form. We are almost done. Now, we need to fix the PHI nodes
+ // that we vectorized. The PHI nodes are currently empty because we did
+ // not want to introduce cycles. Notice that the remaining PHI nodes
+ // that we need to fix are reduction variables.
- const SCEVAddRecExpr *S = dyn_cast<SCEVAddRecExpr>(V);
- if (!S)
- return nullptr;
+ // Create the 'reduced' values for each of the induction vars.
+ // The reduced values are the vector values that we scalarize and combine
+ // after the loop is finished.
+ for (PhiVector::iterator it = RdxPHIsToFix.begin(), e = RdxPHIsToFix.end();
+ it != e; ++it) {
+ PHINode *RdxPhi = *it;
+ assert(RdxPhi && "Unable to recover vectorized PHI");
- V = S->getStepRecurrence(*SE);
- if (!V)
- return nullptr;
+ // Find the reduction variable descriptor.
+ assert(Legal->getReductionVars()->count(RdxPhi) &&
+ "Unable to find the reduction variable");
+ RecurrenceDescriptor RdxDesc = (*Legal->getReductionVars())[RdxPhi];
- // Strip off the size of access multiplication if we are still analyzing the
- // pointer.
- if (OrigPtr == Ptr) {
- DL->getTypeAllocSize(PtrTy->getElementType());
- if (const SCEVMulExpr *M = dyn_cast<SCEVMulExpr>(V)) {
- if (M->getOperand(0)->getSCEVType() != scConstant)
- return nullptr;
+ RecurrenceDescriptor::RecurrenceKind RK = RdxDesc.getRecurrenceKind();
+ TrackingVH<Value> ReductionStartValue = RdxDesc.getRecurrenceStartValue();
+ Instruction *LoopExitInst = RdxDesc.getLoopExitInstr();
+ RecurrenceDescriptor::MinMaxRecurrenceKind MinMaxKind =
+ RdxDesc.getMinMaxRecurrenceKind();
+ setDebugLocFromInst(Builder, ReductionStartValue);
- const APInt &APStepVal =
- cast<SCEVConstant>(M->getOperand(0))->getValue()->getValue();
+ // We need to generate a reduction vector from the incoming scalar.
+ // To do so, we need to generate the 'identity' vector and override
+ // one of the elements with the incoming scalar reduction. We need
+ // to do it in the vector-loop preheader.
+ Builder.SetInsertPoint(LoopBypassBlocks[1]->getTerminator());
- // Huge step value - give up.
- if (APStepVal.getBitWidth() > 64)
- return nullptr;
+ // This is the vector-clone of the value that leaves the loop.
+ VectorParts &VectorExit = getVectorValue(LoopExitInst);
+ Type *VecTy = VectorExit[0]->getType();
- int64_t StepVal = APStepVal.getSExtValue();
- if (PtrAccessSize != StepVal)
- return nullptr;
- V = M->getOperand(1);
+ // Find the reduction identity variable. Zero for addition, or, xor,
+ // one for multiplication, -1 for And.
+ Value *Identity;
+ Value *VectorStart;
+ if (RK == RecurrenceDescriptor::RK_IntegerMinMax ||
+ RK == RecurrenceDescriptor::RK_FloatMinMax) {
+ // MinMax reduction have the start value as their identify.
+ if (VF == 1) {
+ VectorStart = Identity = ReductionStartValue;
+ } else {
+ VectorStart = Identity =
+ Builder.CreateVectorSplat(VF, ReductionStartValue, "minmax.ident");
+ }
+ } else {
+ // Handle other reduction kinds:
+ Constant *Iden = RecurrenceDescriptor::getRecurrenceIdentity(
+ RK, VecTy->getScalarType());
+ if (VF == 1) {
+ Identity = Iden;
+ // This vector is the Identity vector where the first element is the
+ // incoming scalar reduction.
+ VectorStart = ReductionStartValue;
+ } else {
+ Identity = ConstantVector::getSplat(VF, Iden);
+
+ // This vector is the Identity vector where the first element is the
+ // incoming scalar reduction.
+ VectorStart =
+ Builder.CreateInsertElement(Identity, ReductionStartValue, Zero);
+ }
}
- }
- // Strip off casts.
- Type *StripedOffRecurrenceCast = nullptr;
- if (const SCEVCastExpr *C = dyn_cast<SCEVCastExpr>(V)) {
- StripedOffRecurrenceCast = C->getType();
- V = C->getOperand();
- }
+ // Fix the vector-loop phi.
- // Look for the loop invariant symbolic value.
- const SCEVUnknown *U = dyn_cast<SCEVUnknown>(V);
- if (!U)
- return nullptr;
+ // Reductions do not have to start at zero. They can start with
+ // any loop invariant values.
+ VectorParts &VecRdxPhi = WidenMap.get(RdxPhi);
+ BasicBlock *Latch = OrigLoop->getLoopLatch();
+ Value *LoopVal = RdxPhi->getIncomingValueForBlock(Latch);
+ VectorParts &Val = getVectorValue(LoopVal);
+ for (unsigned part = 0; part < UF; ++part) {
+ // Make sure to add the reduction stat value only to the
+ // first unroll part.
+ Value *StartVal = (part == 0) ? VectorStart : Identity;
+ cast<PHINode>(VecRdxPhi[part])->addIncoming(StartVal,
+ LoopVectorPreHeader);
+ cast<PHINode>(VecRdxPhi[part])->addIncoming(Val[part],
+ LoopVectorBody.back());
+ }
+
+ // Before each round, move the insertion point right between
+ // the PHIs and the values we are going to write.
+ // This allows us to write both PHINodes and the extractelement
+ // instructions.
+ Builder.SetInsertPoint(LoopMiddleBlock->getFirstInsertionPt());
+
+ VectorParts RdxParts;
+ setDebugLocFromInst(Builder, LoopExitInst);
+ for (unsigned part = 0; part < UF; ++part) {
+ // This PHINode contains the vectorized reduction variable, or
+ // the initial value vector, if we bypass the vector loop.
+ VectorParts &RdxExitVal = getVectorValue(LoopExitInst);
+ PHINode *NewPhi = Builder.CreatePHI(VecTy, 2, "rdx.vec.exit.phi");
+ Value *StartVal = (part == 0) ? VectorStart : Identity;
+ for (unsigned I = 1, E = LoopBypassBlocks.size(); I != E; ++I)
+ NewPhi->addIncoming(StartVal, LoopBypassBlocks[I]);
+ NewPhi->addIncoming(RdxExitVal[part],
+ LoopVectorBody.back());
+ RdxParts.push_back(NewPhi);
+ }
+
+ // Reduce all of the unrolled parts into a single vector.
+ Value *ReducedPartRdx = RdxParts[0];
+ unsigned Op = RecurrenceDescriptor::getRecurrenceBinOp(RK);
+ setDebugLocFromInst(Builder, ReducedPartRdx);
+ for (unsigned part = 1; part < UF; ++part) {
+ if (Op != Instruction::ICmp && Op != Instruction::FCmp)
+ // Floating point operations had to be 'fast' to enable the reduction.
+ ReducedPartRdx = addFastMathFlag(
+ Builder.CreateBinOp((Instruction::BinaryOps)Op, RdxParts[part],
+ ReducedPartRdx, "bin.rdx"));
+ else
+ ReducedPartRdx = RecurrenceDescriptor::createMinMaxOp(
+ Builder, MinMaxKind, ReducedPartRdx, RdxParts[part]);
+ }
+
+ if (VF > 1) {
+ // VF is a power of 2 so we can emit the reduction using log2(VF) shuffles
+ // and vector ops, reducing the set of values being computed by half each
+ // round.
+ assert(isPowerOf2_32(VF) &&
+ "Reduction emission only supported for pow2 vectors!");
+ Value *TmpVec = ReducedPartRdx;
+ SmallVector<Constant*, 32> ShuffleMask(VF, nullptr);
+ for (unsigned i = VF; i != 1; i >>= 1) {
+ // Move the upper half of the vector to the lower half.
+ for (unsigned j = 0; j != i/2; ++j)
+ ShuffleMask[j] = Builder.getInt32(i/2 + j);
+
+ // Fill the rest of the mask with undef.
+ std::fill(&ShuffleMask[i/2], ShuffleMask.end(),
+ UndefValue::get(Builder.getInt32Ty()));
+
+ Value *Shuf =
+ Builder.CreateShuffleVector(TmpVec,
+ UndefValue::get(TmpVec->getType()),
+ ConstantVector::get(ShuffleMask),
+ "rdx.shuf");
+
+ if (Op != Instruction::ICmp && Op != Instruction::FCmp)
+ // Floating point operations had to be 'fast' to enable the reduction.
+ TmpVec = addFastMathFlag(Builder.CreateBinOp(
+ (Instruction::BinaryOps)Op, TmpVec, Shuf, "bin.rdx"));
+ else
+ TmpVec = RecurrenceDescriptor::createMinMaxOp(Builder, MinMaxKind,
+ TmpVec, Shuf);
+ }
+
+ // The result is in the first element of the vector.
+ ReducedPartRdx = Builder.CreateExtractElement(TmpVec,
+ Builder.getInt32(0));
+ }
+
+ // Create a phi node that merges control-flow from the backedge-taken check
+ // block and the middle block.
+ PHINode *BCBlockPhi = PHINode::Create(RdxPhi->getType(), 2, "bc.merge.rdx",
+ LoopScalarPreHeader->getTerminator());
+ BCBlockPhi->addIncoming(ReductionStartValue, LoopBypassBlocks[0]);
+ BCBlockPhi->addIncoming(ReducedPartRdx, LoopMiddleBlock);
+
+ // Now, we need to fix the users of the reduction variable
+ // inside and outside of the scalar remainder loop.
+ // We know that the loop is in LCSSA form. We need to update the
+ // PHI nodes in the exit blocks.
+ for (BasicBlock::iterator LEI = LoopExitBlock->begin(),
+ LEE = LoopExitBlock->end(); LEI != LEE; ++LEI) {
+ PHINode *LCSSAPhi = dyn_cast<PHINode>(LEI);
+ if (!LCSSAPhi) break;
+
+ // All PHINodes need to have a single entry edge, or two if
+ // we already fixed them.
+ assert(LCSSAPhi->getNumIncomingValues() < 3 && "Invalid LCSSA PHI");
+
+ // We found our reduction value exit-PHI. Update it with the
+ // incoming bypass edge.
+ if (LCSSAPhi->getIncomingValue(0) == LoopExitInst) {
+ // Add an edge coming from the bypass.
+ LCSSAPhi->addIncoming(ReducedPartRdx, LoopMiddleBlock);
+ break;
+ }
+ }// end of the LCSSA phi scan.
- Value *Stride = U->getValue();
- if (!Lp->isLoopInvariant(Stride))
- return nullptr;
+ // Fix the scalar loop reduction variable with the incoming reduction sum
+ // from the vector body and from the backedge value.
+ int IncomingEdgeBlockIdx =
+ (RdxPhi)->getBasicBlockIndex(OrigLoop->getLoopLatch());
+ assert(IncomingEdgeBlockIdx >= 0 && "Invalid block index");
+ // Pick the other block.
+ int SelfEdgeBlockIdx = (IncomingEdgeBlockIdx ? 0 : 1);
+ (RdxPhi)->setIncomingValue(SelfEdgeBlockIdx, BCBlockPhi);
+ (RdxPhi)->setIncomingValue(IncomingEdgeBlockIdx, LoopExitInst);
+ }// end of for each redux variable.
- // If we have stripped off the recurrence cast we have to make sure that we
- // return the value that is used in this loop so that we can replace it later.
- if (StripedOffRecurrenceCast)
- Stride = getUniqueCastUse(Stride, Lp, StripedOffRecurrenceCast);
+ fixLCSSAPHIs();
- return Stride;
+ // Remove redundant induction instructions.
+ cse(LoopVectorBody);
}
-void LoopVectorizationLegality::collectStridedAccess(Value *MemAccess) {
- Value *Ptr = nullptr;
- if (LoadInst *LI = dyn_cast<LoadInst>(MemAccess))
- Ptr = LI->getPointerOperand();
- else if (StoreInst *SI = dyn_cast<StoreInst>(MemAccess))
- Ptr = SI->getPointerOperand();
- else
- return;
-
- Value *Stride = getStrideFromPointer(Ptr, SE, DL, TheLoop);
- if (!Stride)
- return;
-
- DEBUG(dbgs() << "LV: Found a strided access that we can version");
- DEBUG(dbgs() << " Ptr: " << *Ptr << " Stride: " << *Stride << "\n");
- Strides[Ptr] = Stride;
- StrideSet.insert(Stride);
+void InnerLoopVectorizer::fixLCSSAPHIs() {
+ for (BasicBlock::iterator LEI = LoopExitBlock->begin(),
+ LEE = LoopExitBlock->end(); LEI != LEE; ++LEI) {
+ PHINode *LCSSAPhi = dyn_cast<PHINode>(LEI);
+ if (!LCSSAPhi) break;
+ if (LCSSAPhi->getNumIncomingValues() == 1)
+ LCSSAPhi->addIncoming(UndefValue::get(LCSSAPhi->getType()),
+ LoopMiddleBlock);
+ }
}
-void LoopVectorizationLegality::collectLoopUniforms() {
- // We now know that the loop is vectorizable!
- // Collect variables that will remain uniform after vectorization.
- std::vector<Value*> Worklist;
- BasicBlock *Latch = TheLoop->getLoopLatch();
+InnerLoopVectorizer::VectorParts
+InnerLoopVectorizer::createEdgeMask(BasicBlock *Src, BasicBlock *Dst) {
+ assert(std::find(pred_begin(Dst), pred_end(Dst), Src) != pred_end(Dst) &&
+ "Invalid edge");
- // Start with the conditional branch and walk up the block.
- Worklist.push_back(Latch->getTerminator()->getOperand(0));
+ // Look for cached value.
+ std::pair<BasicBlock*, BasicBlock*> Edge(Src, Dst);
+ EdgeMaskCache::iterator ECEntryIt = MaskCache.find(Edge);
+ if (ECEntryIt != MaskCache.end())
+ return ECEntryIt->second;
- // Also add all consecutive pointer values; these values will be uniform
- // after vectorization (and subsequent cleanup) and, until revectorization is
- // supported, all dependencies must also be uniform.
- for (Loop::block_iterator B = TheLoop->block_begin(),
- BE = TheLoop->block_end(); B != BE; ++B)
- for (BasicBlock::iterator I = (*B)->begin(), IE = (*B)->end();
- I != IE; ++I)
- if (I->getType()->isPointerTy() && isConsecutivePtr(I))
- Worklist.insert(Worklist.end(), I->op_begin(), I->op_end());
+ VectorParts SrcMask = createBlockInMask(Src);
- while (!Worklist.empty()) {
- Instruction *I = dyn_cast<Instruction>(Worklist.back());
- Worklist.pop_back();
+ // The terminator has to be a branch inst!
+ BranchInst *BI = dyn_cast<BranchInst>(Src->getTerminator());
+ assert(BI && "Unexpected terminator found");
- // Look at instructions inside this loop.
- // Stop when reaching PHI nodes.
- // TODO: we need to follow values all over the loop, not only in this block.
- if (!I || !TheLoop->contains(I) || isa<PHINode>(I))
- continue;
+ if (BI->isConditional()) {
+ VectorParts EdgeMask = getVectorValue(BI->getCondition());
- // This is a known uniform.
- Uniforms.insert(I);
+ if (BI->getSuccessor(0) != Dst)
+ for (unsigned part = 0; part < UF; ++part)
+ EdgeMask[part] = Builder.CreateNot(EdgeMask[part]);
- // Insert all operands.
- Worklist.insert(Worklist.end(), I->op_begin(), I->op_end());
- }
-}
+ for (unsigned part = 0; part < UF; ++part)
+ EdgeMask[part] = Builder.CreateAnd(EdgeMask[part], SrcMask[part]);
-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;
+ MaskCache[Edge] = EdgeMask;
+ return EdgeMask;
+ }
- /// \brief Set of potential dependent memory accesses.
- typedef EquivalenceClasses<MemAccessInfo> DepCandidates;
+ MaskCache[Edge] = SrcMask;
+ return SrcMask;
+}
- AccessAnalysis(const DataLayout *Dl, AliasAnalysis *AA, DepCandidates &DA) :
- DL(Dl), AST(*AA), DepCands(DA), IsRTCheckNeeded(false) {}
+InnerLoopVectorizer::VectorParts
+InnerLoopVectorizer::createBlockInMask(BasicBlock *BB) {
+ assert(OrigLoop->contains(BB) && "Block is not a part of a loop");
- /// \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);
+ // Loop incoming mask is all-one.
+ if (OrigLoop->getHeader() == BB) {
+ Value *C = ConstantInt::get(IntegerType::getInt1Ty(BB->getContext()), 1);
+ return getVectorValue(C);
}
- /// \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));
+ // This is the block mask. We OR all incoming edges, and with zero.
+ Value *Zero = ConstantInt::get(IntegerType::getInt1Ty(BB->getContext()), 0);
+ VectorParts BlockMask = getVectorValue(Zero);
+
+ // For each pred:
+ for (pred_iterator it = pred_begin(BB), e = pred_end(BB); it != e; ++it) {
+ VectorParts EM = createEdgeMask(*it, BB);
+ for (unsigned part = 0; part < UF; ++part)
+ BlockMask[part] = Builder.CreateOr(BlockMask[part], EM[part]);
}
- /// \brief Check whether we can check the pointers at runtime for
- /// non-intersection.
- bool canCheckPtrAtRT(RuntimePointerCheck &RtCheck, unsigned &NumComparisons,
- ScalarEvolution *SE, Loop *TheLoop,
- ValueToValueMap &Strides,
- bool ShouldCheckStride = false);
+ return BlockMask;
+}
- /// \brief Goes over all memory accesses, checks whether a RT check is needed
- /// and builds sets of dependent accesses.
- void buildDependenceSets() {
- processMemAccesses();
+void InnerLoopVectorizer::widenPHIInstruction(Instruction *PN,
+ InnerLoopVectorizer::VectorParts &Entry,
+ unsigned UF, unsigned VF, PhiVector *PV) {
+ PHINode* P = cast<PHINode>(PN);
+ // Handle reduction variables:
+ if (Legal->getReductionVars()->count(P)) {
+ for (unsigned part = 0; part < UF; ++part) {
+ // This is phase one of vectorizing PHIs.
+ Type *VecTy = (VF == 1) ? PN->getType() :
+ VectorType::get(PN->getType(), VF);
+ Entry[part] = PHINode::Create(VecTy, 2, "vec.phi",
+ LoopVectorBody.back()-> getFirstInsertionPt());
+ }
+ PV->push_back(P);
+ return;
}
- bool isRTCheckNeeded() { return IsRTCheckNeeded; }
+ setDebugLocFromInst(Builder, P);
+ // Check for PHI nodes that are lowered to vector selects.
+ if (P->getParent() != OrigLoop->getHeader()) {
+ // We know that all PHIs in non-header blocks are converted into
+ // selects, so we don't have to worry about the insertion order and we
+ // can just use the builder.
+ // At this point we generate the predication tree. There may be
+ // duplications since this is a simple recursive scan, but future
+ // optimizations will clean it up.
- bool isDependencyCheckNeeded() { return !CheckDeps.empty(); }
- void resetDepChecks() { CheckDeps.clear(); }
+ unsigned NumIncoming = P->getNumIncomingValues();
- MemAccessInfoSet &getDependenciesToCheck() { return CheckDeps; }
+ // Generate a sequence of selects of the form:
+ // SELECT(Mask3, In3,
+ // SELECT(Mask2, In2,
+ // ( ...)))
+ for (unsigned In = 0; In < NumIncoming; In++) {
+ VectorParts Cond = createEdgeMask(P->getIncomingBlock(In),
+ P->getParent());
+ VectorParts &In0 = getVectorValue(P->getIncomingValue(In));
-private:
- typedef SetVector<MemAccessInfo> PtrAccessSet;
+ for (unsigned part = 0; part < UF; ++part) {
+ // We might have single edge PHIs (blocks) - use an identity
+ // 'select' for the first PHI operand.
+ if (In == 0)
+ Entry[part] = Builder.CreateSelect(Cond[part], In0[part],
+ In0[part]);
+ else
+ // Select between the current value and the previous incoming edge
+ // based on the incoming mask.
+ Entry[part] = Builder.CreateSelect(Cond[part], In0[part],
+ Entry[part], "predphi");
+ }
+ }
+ return;
+ }
+
+ // This PHINode must be an induction variable.
+ // Make sure that we know about it.
+ assert(Legal->getInductionVars()->count(P) &&
+ "Not an induction variable");
- /// \brief Go over all memory access and check whether runtime pointer checks
- /// are needed /// and build sets of dependency check candidates.
- void processMemAccesses();
+ LoopVectorizationLegality::InductionInfo II =
+ Legal->getInductionVars()->lookup(P);
- /// Set of all accesses.
- PtrAccessSet Accesses;
+ // FIXME: The newly created binary instructions should contain nsw/nuw flags,
+ // which can be found from the original scalar operations.
+ switch (II.IK) {
+ case LoopVectorizationLegality::IK_NoInduction:
+ llvm_unreachable("Unknown induction");
+ case LoopVectorizationLegality::IK_IntInduction: {
+ assert(P->getType() == II.StartValue->getType() && "Types must match");
+ Type *PhiTy = P->getType();
+ Value *Broadcasted;
+ if (P == OldInduction) {
+ // Handle the canonical induction variable. We might have had to
+ // extend the type.
+ Broadcasted = Builder.CreateTrunc(Induction, PhiTy);
+ } else {
+ // Handle other induction variables that are now based on the
+ // canonical one.
+ Value *NormalizedIdx = Builder.CreateSub(Induction, ExtendedIdx,
+ "normalized.idx");
+ NormalizedIdx = Builder.CreateSExtOrTrunc(NormalizedIdx, PhiTy);
+ Broadcasted = II.transform(Builder, NormalizedIdx);
+ Broadcasted->setName("offset.idx");
+ }
+ Broadcasted = getBroadcastInstrs(Broadcasted);
+ // After broadcasting the induction variable we need to make the vector
+ // consecutive by adding 0, 1, 2, etc.
+ for (unsigned part = 0; part < UF; ++part)
+ Entry[part] = getStepVector(Broadcasted, VF * part, II.StepValue);
+ return;
+ }
+ case LoopVectorizationLegality::IK_PtrInduction:
+ // Handle the pointer induction variable case.
+ assert(P->getType()->isPointerTy() && "Unexpected type.");
+ // This is the normalized GEP that starts counting at zero.
+ Value *NormalizedIdx =
+ Builder.CreateSub(Induction, ExtendedIdx, "normalized.idx");
+ NormalizedIdx =
+ Builder.CreateSExtOrTrunc(NormalizedIdx, II.StepValue->getType());
+ // This is the vector of results. Notice that we don't generate
+ // vector geps because scalar geps result in better code.
+ for (unsigned part = 0; part < UF; ++part) {
+ if (VF == 1) {
+ int EltIndex = part;
+ Constant *Idx = ConstantInt::get(NormalizedIdx->getType(), EltIndex);
+ Value *GlobalIdx = Builder.CreateAdd(NormalizedIdx, Idx);
+ Value *SclrGep = II.transform(Builder, GlobalIdx);
+ SclrGep->setName("next.gep");
+ Entry[part] = SclrGep;
+ continue;
+ }
- /// Set of accesses that need a further dependence check.
- MemAccessInfoSet CheckDeps;
+ Value *VecVal = UndefValue::get(VectorType::get(P->getType(), VF));
+ for (unsigned int i = 0; i < VF; ++i) {
+ int EltIndex = i + part * VF;
+ Constant *Idx = ConstantInt::get(NormalizedIdx->getType(), EltIndex);
+ Value *GlobalIdx = Builder.CreateAdd(NormalizedIdx, Idx);
+ Value *SclrGep = II.transform(Builder, GlobalIdx);
+ SclrGep->setName("next.gep");
+ VecVal = Builder.CreateInsertElement(VecVal, SclrGep,
+ Builder.getInt32(i),
+ "insert.gep");
+ }
+ Entry[part] = VecVal;
+ }
+ return;
+ }
+}
- /// Set of pointers that are read only.
- SmallPtrSet<Value*, 16> ReadOnlyPtr;
+void InnerLoopVectorizer::vectorizeBlockInLoop(BasicBlock *BB, PhiVector *PV) {
+ // For each instruction in the old loop.
+ for (BasicBlock::iterator it = BB->begin(), e = BB->end(); it != e; ++it) {
+ VectorParts &Entry = WidenMap.get(it);
+ switch (it->getOpcode()) {
+ case Instruction::Br:
+ // Nothing to do for PHIs and BR, since we already took care of the
+ // loop control flow instructions.
+ continue;
+ case Instruction::PHI: {
+ // Vectorize PHINodes.
+ widenPHIInstruction(it, Entry, UF, VF, PV);
+ continue;
+ }// End of PHI.
- const DataLayout *DL;
+ case Instruction::Add:
+ case Instruction::FAdd:
+ case Instruction::Sub:
+ case Instruction::FSub:
+ case Instruction::Mul:
+ case Instruction::FMul:
+ case Instruction::UDiv:
+ case Instruction::SDiv:
+ case Instruction::FDiv:
+ case Instruction::URem:
+ case Instruction::SRem:
+ case Instruction::FRem:
+ case Instruction::Shl:
+ case Instruction::LShr:
+ case Instruction::AShr:
+ case Instruction::And:
+ case Instruction::Or:
+ case Instruction::Xor: {
+ // Just widen binops.
+ BinaryOperator *BinOp = dyn_cast<BinaryOperator>(it);
+ setDebugLocFromInst(Builder, BinOp);
+ VectorParts &A = getVectorValue(it->getOperand(0));
+ VectorParts &B = getVectorValue(it->getOperand(1));
- /// An alias set tracker to partition the access set by underlying object and
- //intrinsic property (such as TBAA metadata).
- AliasSetTracker AST;
+ // Use this vector value for all users of the original instruction.
+ for (unsigned Part = 0; Part < UF; ++Part) {
+ Value *V = Builder.CreateBinOp(BinOp->getOpcode(), A[Part], B[Part]);
- /// 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;
+ if (BinaryOperator *VecOp = dyn_cast<BinaryOperator>(V))
+ VecOp->copyIRFlags(BinOp);
- bool IsRTCheckNeeded;
-};
+ Entry[Part] = V;
+ }
-} // end anonymous namespace
+ propagateMetadata(Entry, it);
+ break;
+ }
+ case Instruction::Select: {
+ // Widen selects.
+ // If the selector is loop invariant we can create a select
+ // instruction with a scalar condition. Otherwise, use vector-select.
+ bool InvariantCond = SE->isLoopInvariant(SE->getSCEV(it->getOperand(0)),
+ OrigLoop);
+ setDebugLocFromInst(Builder, it);
-/// \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;
+ // The condition can be loop invariant but still defined inside the
+ // loop. This means that we can't just use the original 'cond' value.
+ // We have to take the 'vectorized' value and pick the first lane.
+ // Instcombine will make this a no-op.
+ VectorParts &Cond = getVectorValue(it->getOperand(0));
+ VectorParts &Op0 = getVectorValue(it->getOperand(1));
+ VectorParts &Op1 = getVectorValue(it->getOperand(2));
- return AR->isAffine();
-}
+ Value *ScalarCond = (VF == 1) ? Cond[0] :
+ Builder.CreateExtractElement(Cond[0], Builder.getInt32(0));
-/// \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(
- 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;
+ for (unsigned Part = 0; Part < UF; ++Part) {
+ Entry[Part] = Builder.CreateSelect(
+ InvariantCond ? ScalarCond : Cond[Part],
+ Op0[Part],
+ Op1[Part]);
}
- }
- if (IsDepCheckNeeded && CanDoRT && RunningDepId == 2)
- NumComparisons += 0; // Only one dependence set.
- else {
- NumComparisons += (NumWritePtrChecks * (NumReadPtrChecks +
- NumWritePtrChecks - 1));
+ propagateMetadata(Entry, it);
+ break;
}
- ++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;
+ case Instruction::ICmp:
+ case Instruction::FCmp: {
+ // Widen compares. Generate vector compares.
+ bool FCmp = (it->getOpcode() == Instruction::FCmp);
+ CmpInst *Cmp = dyn_cast<CmpInst>(it);
+ setDebugLocFromInst(Builder, it);
+ VectorParts &A = getVectorValue(it->getOperand(0));
+ VectorParts &B = getVectorValue(it->getOperand(1));
+ for (unsigned Part = 0; Part < UF; ++Part) {
+ Value *C = nullptr;
+ if (FCmp)
+ C = Builder.CreateFCmp(Cmp->getPredicate(), A[Part], B[Part]);
+ else
+ C = Builder.CreateICmp(Cmp->getPredicate(), A[Part], B[Part]);
+ Entry[Part] = C;
+ }
- Value *PtrI = RtCheck.Pointers[i];
- Value *PtrJ = RtCheck.Pointers[j];
+ propagateMetadata(Entry, it);
+ break;
+ }
- 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;
+ case Instruction::Store:
+ case Instruction::Load:
+ vectorizeMemoryInstruction(it);
+ break;
+ case Instruction::ZExt:
+ case Instruction::SExt:
+ case Instruction::FPToUI:
+ case Instruction::FPToSI:
+ case Instruction::FPExt:
+ case Instruction::PtrToInt:
+ case Instruction::IntToPtr:
+ case Instruction::SIToFP:
+ case Instruction::UIToFP:
+ case Instruction::Trunc:
+ case Instruction::FPTrunc:
+ case Instruction::BitCast: {
+ CastInst *CI = dyn_cast<CastInst>(it);
+ setDebugLocFromInst(Builder, it);
+ /// Optimize the special case where the source is the induction
+ /// variable. Notice that we can only optimize the 'trunc' case
+ /// because: a. FP conversions lose precision, b. sext/zext may wrap,
+ /// c. other casts depend on pointer size.
+ if (CI->getOperand(0) == OldInduction &&
+ it->getOpcode() == Instruction::Trunc) {
+ Value *ScalarCast = Builder.CreateCast(CI->getOpcode(), Induction,
+ CI->getType());
+ Value *Broadcasted = getBroadcastInstrs(ScalarCast);
+ LoopVectorizationLegality::InductionInfo II =
+ Legal->getInductionVars()->lookup(OldInduction);
+ Constant *Step =
+ ConstantInt::getSigned(CI->getType(), II.StepValue->getSExtValue());
+ for (unsigned Part = 0; Part < UF; ++Part)
+ Entry[Part] = getStepVector(Broadcasted, VF * Part, Step);
+ propagateMetadata(Entry, it);
+ break;
}
+ /// Vectorize casts.
+ Type *DestTy = (VF == 1) ? CI->getType() :
+ VectorType::get(CI->getType(), VF);
+
+ VectorParts &A = getVectorValue(it->getOperand(0));
+ for (unsigned Part = 0; Part < UF; ++Part)
+ Entry[Part] = Builder.CreateCast(CI->getOpcode(), A[Part], DestTy);
+ propagateMetadata(Entry, it);
+ break;
}
- }
- return CanDoRT;
-}
+ case Instruction::Call: {
+ // Ignore dbg intrinsics.
+ if (isa<DbgInfoIntrinsic>(it))
+ break;
+ setDebugLocFromInst(Builder, it);
-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 AV : AS) {
- Value *Ptr = AV.getValue();
-
- // For a single memory access in AliasSetTracker, Accesses may contain
- // both read and write, and they both need to be handled for CheckDeps.
- for (auto AC : S) {
- if (AC.getPointer() != Ptr)
- continue;
+ Module *M = BB->getParent()->getParent();
+ CallInst *CI = cast<CallInst>(it);
- bool IsWrite = AC.getInt();
+ StringRef FnName = CI->getCalledFunction()->getName();
+ Function *F = CI->getCalledFunction();
+ Type *RetTy = ToVectorTy(CI->getType(), VF);
+ SmallVector<Type *, 4> Tys;
+ for (unsigned i = 0, ie = CI->getNumArgOperands(); i != ie; ++i)
+ Tys.push_back(ToVectorTy(CI->getArgOperand(i)->getType(), VF));
- // 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;
- }
+ Intrinsic::ID ID = getIntrinsicIDForCall(CI, TLI);
+ if (ID &&
+ (ID == Intrinsic::assume || ID == Intrinsic::lifetime_end ||
+ ID == Intrinsic::lifetime_start)) {
+ scalarizeInstruction(it);
+ break;
+ }
+ // The flag shows whether we use Intrinsic or a usual Call for vectorized
+ // version of the instruction.
+ // Is it beneficial to perform intrinsic call compared to lib call?
+ bool NeedToScalarize;
+ unsigned CallCost = getVectorCallCost(CI, VF, *TTI, TLI, NeedToScalarize);
+ bool UseVectorIntrinsic =
+ ID && getVectorIntrinsicCost(CI, VF, *TTI, TLI) <= CallCost;
+ if (!UseVectorIntrinsic && NeedToScalarize) {
+ scalarizeInstruction(it);
+ break;
+ }
- // 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;
+ for (unsigned Part = 0; Part < UF; ++Part) {
+ SmallVector<Value *, 4> Args;
+ for (unsigned i = 0, ie = CI->getNumArgOperands(); i != ie; ++i) {
+ Value *Arg = CI->getArgOperand(i);
+ // Some intrinsics have a scalar argument - don't replace it with a
+ // vector.
+ if (!UseVectorIntrinsic || !hasVectorInstrinsicScalarOpd(ID, i)) {
+ VectorParts &VectorArg = getVectorValue(CI->getArgOperand(i));
+ Arg = VectorArg[Part];
}
+ Args.push_back(Arg);
+ }
- 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;
+ Function *VectorF;
+ if (UseVectorIntrinsic) {
+ // Use vector version of the intrinsic.
+ Type *TysForDecl[] = {CI->getType()};
+ if (VF > 1)
+ TysForDecl[0] = VectorType::get(CI->getType()->getScalarType(), VF);
+ VectorF = Intrinsic::getDeclaration(M, ID, TysForDecl);
+ } else {
+ // Use vector version of the library call.
+ StringRef VFnName = TLI->getVectorizedFunction(FnName, VF);
+ assert(!VFnName.empty() && "Vector function name is empty.");
+ VectorF = M->getFunction(VFnName);
+ if (!VectorF) {
+ // Generate a declaration
+ FunctionType *FTy = FunctionType::get(RetTy, Tys, false);
+ VectorF =
+ Function::Create(FTy, Function::ExternalLinkage, VFnName, M);
+ VectorF->copyAttributesFrom(F);
}
}
+ assert(VectorF && "Can't create vector function.");
+ Entry[Part] = Builder.CreateCall(VectorF, Args);
}
- }
- }
-}
-
-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,
- const LoopAccessAnalysis::VectorizerParams &VectParams)
- : SE(Se), DL(Dl), InnermostLoop(L), AccessIdx(0),
- ShouldRetryWithRuntimeCheck(false), VectParams(VectParams) {}
-
- /// \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 Vectorizer parameters used by the analysis.
- LoopAccessAnalysis::VectorizerParams VectParams;
- /// \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
+ propagateMetadata(Entry, it);
+ break;
+ }
-static bool isInBoundsGep(Value *Ptr) {
- if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Ptr))
- return GEP->isInBounds();
- return false;
+ default:
+ // All other instructions are unsupported. Scalarize them.
+ scalarizeInstruction(it);
+ break;
+ }// end of switch.
+ }// end of for_each instr.
}
-/// \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;
- }
+void InnerLoopVectorizer::updateAnalysis() {
+ // Forget the original basic block.
+ SE->forgetLoop(OrigLoop);
- // 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;
- }
+ // Update the dominator tree information.
+ assert(DT->properlyDominates(LoopBypassBlocks.front(), LoopExitBlock) &&
+ "Entry does not dominate exit.");
- // Check the step is constant.
- const SCEV *Step = AR->getStepRecurrence(*SE);
+ for (unsigned I = 1, E = LoopBypassBlocks.size(); I != E; ++I)
+ DT->addNewBlock(LoopBypassBlocks[I], LoopBypassBlocks[I-1]);
+ DT->addNewBlock(LoopVectorPreHeader, LoopBypassBlocks.back());
- // 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;
+ // Due to if predication of stores we might create a sequence of "if(pred)
+ // a[i] = ...; " blocks.
+ for (unsigned i = 0, e = LoopVectorBody.size(); i != e; ++i) {
+ if (i == 0)
+ DT->addNewBlock(LoopVectorBody[0], LoopVectorPreHeader);
+ else if (isPredicatedBlock(i)) {
+ DT->addNewBlock(LoopVectorBody[i], LoopVectorBody[i-1]);
+ } else {
+ DT->addNewBlock(LoopVectorBody[i], LoopVectorBody[i-2]);
+ }
}
- 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;
+ DT->addNewBlock(LoopMiddleBlock, LoopBypassBlocks[1]);
+ DT->addNewBlock(LoopScalarPreHeader, LoopBypassBlocks[0]);
+ DT->changeImmediateDominator(LoopScalarBody, LoopScalarPreHeader);
+ DT->changeImmediateDominator(LoopExitBlock, LoopBypassBlocks[0]);
- return Stride;
+ DEBUG(DT->verifyDomTree());
}
-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 = VectParams.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;
+/// \brief Check whether it is safe to if-convert this phi node.
+///
+/// Phi nodes with constant expressions that can trap are not safe to if
+/// convert.
+static bool canIfConvertPHINodes(BasicBlock *BB) {
+ for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
+ PHINode *Phi = dyn_cast<PHINode>(I);
+ if (!Phi)
+ return true;
+ for (unsigned p = 0, e = Phi->getNumIncomingValues(); p != e; ++p)
+ if (Constant *C = dyn_cast<Constant>(Phi->getIncomingValue(p)))
+ if (C->canTrap())
+ return false;
}
-
- if (MaxVFWithoutSLForwardIssues < MaxSafeDepDistBytes &&
- MaxVFWithoutSLForwardIssues != VectParams.MaxVectorWidth*TypeByteSize)
- MaxSafeDepDistBytes = MaxVFWithoutSLForwardIssues;
- return false;
+ return true;
}
-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)
+bool LoopVectorizationLegality::canVectorizeWithIfConvert() {
+ if (!EnableIfConversion) {
+ emitAnalysis(VectorizationReport() << "if-conversion is disabled");
return false;
+ }
- // We cannot check pointers in different address spaces.
- if (APtr->getType()->getPointerAddressSpace() !=
- BPtr->getType()->getPointerAddressSpace())
- return true;
+ assert(TheLoop->getNumBlocks() > 1 && "Single block loops are vectorizable");
- const SCEV *AScev = replaceSymbolicStrideSCEV(SE, Strides, APtr);
- const SCEV *BScev = replaceSymbolicStrideSCEV(SE, Strides, BPtr);
+ // A list of pointers that we can safely read and write to.
+ SmallPtrSet<Value *, 8> SafePointes;
- int StrideAPtr = isStridedPtr(SE, DL, APtr, InnermostLoop, Strides);
- int StrideBPtr = isStridedPtr(SE, DL, BPtr, InnermostLoop, Strides);
+ // Collect safe addresses.
+ for (Loop::block_iterator BI = TheLoop->block_begin(),
+ BE = TheLoop->block_end(); BI != BE; ++BI) {
+ BasicBlock *BB = *BI;
- const SCEV *Src = AScev;
- const SCEV *Sink = BScev;
+ if (blockNeedsPredication(BB))
+ continue;
- // If the induction step is negative we have to invert source and sink of the
- // dependence.
- if (StrideAPtr < 0) {
- //Src = BScev;
- //Sink = AScev;
- std::swap(APtr, BPtr);
- std::swap(Src, Sink);
- std::swap(AIsWrite, BIsWrite);
- std::swap(AIdx, BIdx);
- std::swap(StrideAPtr, StrideBPtr);
+ for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
+ if (LoadInst *LI = dyn_cast<LoadInst>(I))
+ SafePointes.insert(LI->getPointerOperand());
+ else if (StoreInst *SI = dyn_cast<StoreInst>(I))
+ SafePointes.insert(SI->getPointerOperand());
+ }
}
- const SCEV *Dist = SE->getMinusSCEV(Sink, Src);
+ // Collect the blocks that need predication.
+ BasicBlock *Header = TheLoop->getHeader();
+ for (Loop::block_iterator BI = TheLoop->block_begin(),
+ BE = TheLoop->block_end(); BI != BE; ++BI) {
+ BasicBlock *BB = *BI;
- DEBUG(dbgs() << "LV: Src Scev: " << *Src << "Sink Scev: " << *Sink
- << "(Induction step: " << StrideAPtr << ")\n");
- DEBUG(dbgs() << "LV: Distance for " << *InstMap[AIdx] << " to "
- << *InstMap[BIdx] << ": " << *Dist << "\n");
+ // We don't support switch statements inside loops.
+ if (!isa<BranchInst>(BB->getTerminator())) {
+ emitAnalysis(VectorizationReport(BB->getTerminator())
+ << "loop contains a switch statement");
+ return false;
+ }
- // Need consecutive accesses. We don't want to vectorize
- // "A[B[i]] += ..." and similar code or pointer arithmetic that could wrap in
- // the address space.
- if (!StrideAPtr || !StrideBPtr || StrideAPtr != StrideBPtr){
- DEBUG(dbgs() << "Non-consecutive pointer access\n");
- return true;
+ // We must be able to predicate all blocks that need to be predicated.
+ if (blockNeedsPredication(BB)) {
+ if (!blockCanBePredicated(BB, SafePointes)) {
+ emitAnalysis(VectorizationReport(BB->getTerminator())
+ << "control flow cannot be substituted for a select");
+ return false;
+ }
+ } else if (BB != Header && !canIfConvertPHINodes(BB)) {
+ emitAnalysis(VectorizationReport(BB->getTerminator())
+ << "control flow cannot be substituted for a select");
+ return false;
+ }
}
- const SCEVConstant *C = dyn_cast<SCEVConstant>(Dist);
- if (!C) {
- DEBUG(dbgs() << "LV: Dependence because of non-constant distance\n");
- ShouldRetryWithRuntimeCheck = true;
- return true;
+ // We can if-convert this loop.
+ return true;
+}
+
+bool LoopVectorizationLegality::canVectorize() {
+ // We must have a loop in canonical form. Loops with indirectbr in them cannot
+ // be canonicalized.
+ if (!TheLoop->getLoopPreheader()) {
+ emitAnalysis(
+ VectorizationReport() <<
+ "loop control flow is not understood by vectorizer");
+ return false;
}
- Type *ATy = APtr->getType()->getPointerElementType();
- Type *BTy = BPtr->getType()->getPointerElementType();
- unsigned TypeByteSize = DL->getTypeAllocSize(ATy);
+ // We can only vectorize innermost loops.
+ if (!TheLoop->empty()) {
+ emitAnalysis(VectorizationReport() << "loop is not the innermost loop");
+ return false;
+ }
- // Negative distances are not plausible dependencies.
- const APInt &Val = C->getValue()->getValue();
- if (Val.isNegative()) {
- bool IsTrueDataDependence = (AIsWrite && !BIsWrite);
- if (IsTrueDataDependence &&
- (couldPreventStoreLoadForward(Val.abs().getZExtValue(), TypeByteSize) ||
- ATy != BTy))
- return true;
+ // We must have a single backedge.
+ if (TheLoop->getNumBackEdges() != 1) {
+ emitAnalysis(
+ VectorizationReport() <<
+ "loop control flow is not understood by vectorizer");
+ return false;
+ }
- DEBUG(dbgs() << "LV: Dependence is negative: NoDep\n");
+ // We must have a single exiting block.
+ if (!TheLoop->getExitingBlock()) {
+ emitAnalysis(
+ VectorizationReport() <<
+ "loop control flow is not understood by vectorizer");
return false;
}
- // Write to the same location with the same size.
- // Could be improved to assert type sizes are the same (i32 == float, etc).
- if (Val == 0) {
- if (ATy == BTy)
- return false;
- DEBUG(dbgs() << "LV: Zero dependence difference but different types\n");
- return true;
+ // We only handle bottom-tested loops, i.e. loop in which the condition is
+ // checked at the end of each iteration. With that we can assume that all
+ // instructions in the loop are executed the same number of times.
+ if (TheLoop->getExitingBlock() != TheLoop->getLoopLatch()) {
+ emitAnalysis(
+ VectorizationReport() <<
+ "loop control flow is not understood by vectorizer");
+ return false;
}
- assert(Val.isStrictlyPositive() && "Expect a positive value");
+ // We need to have a loop header.
+ DEBUG(dbgs() << "LV: Found a loop: " <<
+ TheLoop->getHeader()->getName() << '\n');
- // Positive distance bigger than max vectorization factor.
- if (ATy != BTy) {
- DEBUG(dbgs() <<
- "LV: ReadWrite-Write positive dependency with different types\n");
+ // Check if we can if-convert non-single-bb loops.
+ unsigned NumBlocks = TheLoop->getNumBlocks();
+ if (NumBlocks != 1 && !canVectorizeWithIfConvert()) {
+ DEBUG(dbgs() << "LV: Can't if-convert the loop.\n");
return false;
}
- unsigned Distance = (unsigned) Val.getZExtValue();
+ // ScalarEvolution needs to be able to find the exit count.
+ const SCEV *ExitCount = SE->getBackedgeTakenCount(TheLoop);
+ if (ExitCount == SE->getCouldNotCompute()) {
+ emitAnalysis(VectorizationReport() <<
+ "could not determine number of loop iterations");
+ DEBUG(dbgs() << "LV: SCEV could not compute the loop exit count.\n");
+ return false;
+ }
- // Bail out early if passed-in parameters make vectorization not feasible.
- unsigned ForcedFactor = (VectParams.VectorizationFactor ?
- VectParams.VectorizationFactor : 1);
- unsigned ForcedUnroll = (VectParams.VectorizationInterleave ?
- VectParams.VectorizationInterleave : 1);
+ // Check if we can vectorize the instructions and CFG in this loop.
+ if (!canVectorizeInstrs()) {
+ DEBUG(dbgs() << "LV: Can't vectorize the instructions or CFG\n");
+ return false;
+ }
- // The distance must be bigger than the size needed for a vectorized version
- // of the operation and the size of the vectorized operation must not be
- // bigger than the currrent maximum size.
- if (Distance < 2*TypeByteSize ||
- 2*TypeByteSize > MaxSafeDepDistBytes ||
- Distance < TypeByteSize * ForcedUnroll * ForcedFactor) {
- DEBUG(dbgs() << "LV: Failure because of Positive distance "
- << Val.getSExtValue() << '\n');
- return true;
+ // Go over each instruction and look at memory deps.
+ if (!canVectorizeMemory()) {
+ DEBUG(dbgs() << "LV: Can't vectorize due to memory conflicts\n");
+ return false;
}
- MaxSafeDepDistBytes = Distance < MaxSafeDepDistBytes ?
- Distance : MaxSafeDepDistBytes;
+ // Collect all of the variables that remain uniform after vectorization.
+ collectLoopUniforms();
- bool IsTrueDataDependence = (!AIsWrite && BIsWrite);
- if (IsTrueDataDependence &&
- couldPreventStoreLoadForward(Distance, TypeByteSize))
- return true;
+ DEBUG(dbgs() << "LV: We can vectorize this loop"
+ << (LAI->getRuntimePointerChecking()->Need
+ ? " (with a runtime bound check)"
+ : "")
+ << "!\n");
- DEBUG(dbgs() << "LV: Positive distance " << Val.getSExtValue() <<
- " with max VF = " << MaxSafeDepDistBytes / TypeByteSize << '\n');
+ bool UseInterleaved = TTI->enableInterleavedAccessVectorization();
- return false;
-}
+ // If an override option has been passed in for interleaved accesses, use it.
+ if (EnableInterleavedMemAccesses.getNumOccurrences() > 0)
+ UseInterleaved = EnableInterleavedMemAccesses;
-bool MemoryDepChecker::areDepsSafe(AccessAnalysis::DepCandidates &AccessSets,
- MemAccessInfoSet &CheckDeps,
- ValueToValueMap &Strides) {
-
- MaxSafeDepDistBytes = -1U;
- while (!CheckDeps.empty()) {
- MemAccessInfo CurAccess = *CheckDeps.begin();
-
- // Get the relevant memory access set.
- EquivalenceClasses<MemAccessInfo>::iterator I =
- AccessSets.findValue(AccessSets.getLeaderValue(CurAccess));
-
- // Check accesses within this set.
- EquivalenceClasses<MemAccessInfo>::member_iterator AI, AE;
- AI = AccessSets.member_begin(I), AE = AccessSets.member_end();
-
- // Check every access pair.
- while (AI != AE) {
- CheckDeps.erase(*AI);
- EquivalenceClasses<MemAccessInfo>::member_iterator OI = std::next(AI);
- while (OI != AE) {
- // Check every accessing instruction pair in program order.
- for (std::vector<unsigned>::iterator I1 = Accesses[*AI].begin(),
- I1E = Accesses[*AI].end(); I1 != I1E; ++I1)
- for (std::vector<unsigned>::iterator I2 = Accesses[*OI].begin(),
- I2E = Accesses[*OI].end(); I2 != I2E; ++I2) {
- if (*I1 < *I2 && isDependent(*AI, *I1, *OI, *I2, Strides))
- return false;
- if (*I2 < *I1 && isDependent(*OI, *I2, *AI, *I1, Strides))
- return false;
- }
- ++OI;
- }
- AI++;
- }
- }
+ // Analyze interleaved memory accesses.
+ if (UseInterleaved)
+ InterleaveInfo.analyzeInterleaving(Strides);
+
+ // Okay! We can vectorize. At this point we don't have any other mem analysis
+ // which may limit our maximum vectorization factor, so just return true with
+ // no restrictions.
return true;
}
-bool LoopAccessAnalysis::canVectorizeMemory(ValueToValueMap &Strides) {
+static Type *convertPointerToIntegerType(const DataLayout &DL, Type *Ty) {
+ if (Ty->isPointerTy())
+ return DL.getIntPtrType(Ty);
+
+ // It is possible that char's or short's overflow when we ask for the loop's
+ // trip count, work around this by changing the type size.
+ if (Ty->getScalarSizeInBits() < 32)
+ return Type::getInt32Ty(Ty->getContext());
- typedef SmallVector<Value*, 16> ValueVector;
- typedef SmallPtrSet<Value*, 16> ValueSet;
+ return Ty;
+}
- // Holds the Load and Store *instructions*.
- ValueVector Loads;
- ValueVector Stores;
+static Type* getWiderType(const DataLayout &DL, Type *Ty0, Type *Ty1) {
+ Ty0 = convertPointerToIntegerType(DL, Ty0);
+ Ty1 = convertPointerToIntegerType(DL, Ty1);
+ if (Ty0->getScalarSizeInBits() > Ty1->getScalarSizeInBits())
+ return Ty0;
+ return Ty1;
+}
- // Holds all the different accesses in the loop.
- unsigned NumReads = 0;
- unsigned NumReadWrites = 0;
+/// \brief Check that the instruction has outside loop users and is not an
+/// identified reduction variable.
+static bool hasOutsideLoopUser(const Loop *TheLoop, Instruction *Inst,
+ SmallPtrSetImpl<Value *> &Reductions) {
+ // Reduction instructions are allowed to have exit users. All other
+ // instructions must not have external users.
+ if (!Reductions.count(Inst))
+ //Check that all of the users of the loop are inside the BB.
+ for (User *U : Inst->users()) {
+ Instruction *UI = cast<Instruction>(U);
+ // This user may be a reduction exit value.
+ if (!TheLoop->contains(UI)) {
+ DEBUG(dbgs() << "LV: Found an outside user for : " << *UI << '\n');
+ return true;
+ }
+ }
+ return false;
+}
- PtrRtCheck.Pointers.clear();
- PtrRtCheck.Need = false;
+bool LoopVectorizationLegality::canVectorizeInstrs() {
+ BasicBlock *PreHeader = TheLoop->getLoopPreheader();
+ BasicBlock *Header = TheLoop->getHeader();
- const bool IsAnnotatedParallel = TheLoop->isAnnotatedParallel();
- MemoryDepChecker DepChecker(SE, DL, TheLoop, VectParams);
+ // Look for the attribute signaling the absence of NaNs.
+ Function &F = *Header->getParent();
+ const DataLayout &DL = F.getParent()->getDataLayout();
+ if (F.hasFnAttribute("no-nans-fp-math"))
+ HasFunNoNaNAttr =
+ F.getFnAttribute("no-nans-fp-math").getValueAsString() == "true";
- // For each block.
+ // For each block in the loop.
for (Loop::block_iterator bb = TheLoop->block_begin(),
be = TheLoop->block_end(); bb != be; ++bb) {
- // Scan the BB and collect legal loads and stores.
+ // Scan the instructions in the block and look for hazards.
for (BasicBlock::iterator it = (*bb)->begin(), e = (*bb)->end(); it != e;
++it) {
- // If this is a load, save it. If this instruction can read from memory
- // but is not a load, then we quit. Notice that we don't handle function
- // calls that read or write.
- if (it->mayReadFromMemory()) {
- // Many math library functions read the rounding mode. We will only
- // vectorize a loop if it contains known function calls that don't set
- // the flag. Therefore, it is safe to ignore this read from memory.
- CallInst *Call = dyn_cast<CallInst>(it);
- if (Call && getIntrinsicIDForCall(Call, TLI))
- continue;
-
- LoadInst *Ld = dyn_cast<LoadInst>(it);
- if (!Ld || (!Ld->isSimple() && !IsAnnotatedParallel)) {
- emitAnalysis(VectorizationReport(Ld)
- << "read with atomic ordering or volatile read");
- DEBUG(dbgs() << "LV: Found a non-simple load.\n");
+ if (PHINode *Phi = dyn_cast<PHINode>(it)) {
+ Type *PhiTy = Phi->getType();
+ // Check that this PHI type is allowed.
+ if (!PhiTy->isIntegerTy() &&
+ !PhiTy->isFloatingPointTy() &&
+ !PhiTy->isPointerTy()) {
+ emitAnalysis(VectorizationReport(it)
+ << "loop control flow is not understood by vectorizer");
+ DEBUG(dbgs() << "LV: Found an non-int non-pointer PHI.\n");
return false;
}
- NumLoads++;
- Loads.push_back(Ld);
- DepChecker.addAccess(Ld);
- continue;
- }
- // Save 'store' instructions. Abort if other instructions write to memory.
- if (it->mayWriteToMemory()) {
- StoreInst *St = dyn_cast<StoreInst>(it);
- if (!St) {
+ // If this PHINode is not in the header block, then we know that we
+ // can convert it to select during if-conversion. No need to check if
+ // the PHIs in this block are induction or reduction variables.
+ if (*bb != Header) {
+ // Check that this instruction has no outside users or is an
+ // identified reduction value with an outside user.
+ if (!hasOutsideLoopUser(TheLoop, it, AllowedExit))
+ continue;
emitAnalysis(VectorizationReport(it) <<
- "instruction cannot be vectorized");
+ "value could not be identified as "
+ "an induction or reduction variable");
return false;
}
- if (!St->isSimple() && !IsAnnotatedParallel) {
- emitAnalysis(VectorizationReport(St)
- << "write with atomic ordering or volatile write");
- DEBUG(dbgs() << "LV: Found a non-simple store.\n");
+
+ // We only allow if-converted PHIs with exactly two incoming values.
+ if (Phi->getNumIncomingValues() != 2) {
+ emitAnalysis(VectorizationReport(it)
+ << "control flow not understood by vectorizer");
+ DEBUG(dbgs() << "LV: Found an invalid PHI.\n");
return false;
}
- NumStores++;
- Stores.push_back(St);
- DepChecker.addAccess(St);
- }
- } // Next instr.
- } // Next block.
-
- // Now we have two lists that hold the loads and the stores.
- // Next, we find the pointers that they use.
-
- // Check if we see any stores. If there are no stores, then we don't
- // care if the pointers are *restrict*.
- if (!Stores.size()) {
- DEBUG(dbgs() << "LV: Found a read-only loop!\n");
- return true;
- }
-
- AccessAnalysis::DepCandidates DependentAccesses;
- AccessAnalysis Accesses(DL, AA, DependentAccesses);
-
- // Holds the analyzed pointers. We don't want to call GetUnderlyingObjects
- // multiple times on the same object. If the ptr is accessed twice, once
- // for read and once for write, it will only appear once (on the write
- // list). This is okay, since we are going to check for conflicts between
- // writes and between reads and writes, but not between reads and reads.
- ValueSet Seen;
-
- ValueVector::iterator I, IE;
- for (I = Stores.begin(), IE = Stores.end(); I != IE; ++I) {
- StoreInst *ST = cast<StoreInst>(*I);
- Value* Ptr = ST->getPointerOperand();
-
- if (isUniform(Ptr)) {
- emitAnalysis(
- VectorizationReport(ST)
- << "write to a loop invariant address could not be vectorized");
- DEBUG(dbgs() << "LV: We don't allow storing to uniform addresses\n");
- return false;
- }
-
- // If we did *not* see this pointer before, insert it to the read-write
- // list. At this phase it is only a 'write' list.
- if (Seen.insert(Ptr).second) {
- ++NumReadWrites;
-
- AliasAnalysis::Location Loc = AA->getLocation(ST);
- // The TBAA metadata could have a control dependency on the predication
- // condition, so we cannot rely on it when determining whether or not we
- // need runtime pointer checks.
- if (blockNeedsPredication(ST->getParent()))
- Loc.AATags.TBAA = nullptr;
-
- Accesses.addStore(Loc);
- }
- }
-
- if (IsAnnotatedParallel) {
- DEBUG(dbgs()
- << "LV: A loop annotated parallel, ignore memory dependency "
- << "checks.\n");
- return true;
- }
-
- for (I = Loads.begin(), IE = Loads.end(); I != IE; ++I) {
- LoadInst *LD = cast<LoadInst>(*I);
- Value* Ptr = LD->getPointerOperand();
- // If we did *not* see this pointer before, insert it to the
- // read list. If we *did* see it before, then it is already in
- // the read-write list. This allows us to vectorize expressions
- // such as A[i] += x; Because the address of A[i] is a read-write
- // pointer. This only works if the index of A[i] is consecutive.
- // If the address of i is unknown (for example A[B[i]]) then we may
- // read a few words, modify, and write a few words, and some of the
- // words may be written to the same address.
- bool IsReadOnlyPtr = false;
- if (Seen.insert(Ptr).second ||
- !isStridedPtr(SE, DL, Ptr, TheLoop, Strides)) {
- ++NumReads;
- IsReadOnlyPtr = true;
- }
-
- AliasAnalysis::Location Loc = AA->getLocation(LD);
- // The TBAA metadata could have a control dependency on the predication
- // condition, so we cannot rely on it when determining whether or not we
- // need runtime pointer checks.
- if (blockNeedsPredication(LD->getParent()))
- Loc.AATags.TBAA = nullptr;
-
- Accesses.addLoad(Loc, IsReadOnlyPtr);
- }
- // If we write (or read-write) to a single destination and there are no
- // other reads in this loop then is it safe to vectorize.
- if (NumReadWrites == 1 && NumReads == 0) {
- DEBUG(dbgs() << "LV: Found a write-only loop!\n");
- return true;
- }
-
- // Build dependence sets and check whether we need a runtime pointer bounds
- // check.
- Accesses.buildDependenceSets();
- bool NeedRTCheck = Accesses.isRTCheckNeeded();
-
- // Find pointers with computable bounds. We are going to use this information
- // to place a runtime bound check.
- unsigned NumComparisons = 0;
- bool CanDoRT = false;
- if (NeedRTCheck)
- CanDoRT = Accesses.canCheckPtrAtRT(PtrRtCheck, NumComparisons, SE, TheLoop,
- Strides);
-
- DEBUG(dbgs() << "LV: We need to do " << NumComparisons <<
- " pointer comparisons.\n");
+ // This is the value coming from the preheader.
+ Value *StartValue = Phi->getIncomingValueForBlock(PreHeader);
+ ConstantInt *StepValue = nullptr;
+ // Check if this is an induction variable.
+ InductionKind IK = isInductionVariable(Phi, StepValue);
- // If we only have one set of dependences to check pointers among we don't
- // need a runtime check.
- if (NumComparisons == 0 && NeedRTCheck)
- NeedRTCheck = false;
+ if (IK_NoInduction != IK) {
+ // Get the widest type.
+ if (!WidestIndTy)
+ WidestIndTy = convertPointerToIntegerType(DL, PhiTy);
+ else
+ WidestIndTy = getWiderType(DL, PhiTy, WidestIndTy);
- // Check that we did not collect too many pointers or found an unsizeable
- // pointer.
- if (!CanDoRT || NumComparisons > VectParams.RuntimeMemoryCheckThreshold) {
- PtrRtCheck.reset();
- CanDoRT = false;
- }
+ // Int inductions are special because we only allow one IV.
+ if (IK == IK_IntInduction && StepValue->isOne()) {
+ // Use the phi node with the widest type as induction. Use the last
+ // one if there are multiple (no good reason for doing this other
+ // than it is expedient).
+ if (!Induction || PhiTy == WidestIndTy)
+ Induction = Phi;
+ }
- if (CanDoRT) {
- DEBUG(dbgs() << "LV: We can perform a memory runtime check if needed.\n");
- }
+ DEBUG(dbgs() << "LV: Found an induction variable.\n");
+ Inductions[Phi] = InductionInfo(StartValue, IK, StepValue);
- if (NeedRTCheck && !CanDoRT) {
- emitAnalysis(VectorizationReport() << "cannot identify array bounds");
- DEBUG(dbgs() << "LV: We can't vectorize because we can't find " <<
- "the array bounds.\n");
- PtrRtCheck.reset();
- return false;
- }
+ // Until we explicitly handle the case of an induction variable with
+ // an outside loop user we have to give up vectorizing this loop.
+ if (hasOutsideLoopUser(TheLoop, it, AllowedExit)) {
+ emitAnalysis(VectorizationReport(it) <<
+ "use of induction value outside of the "
+ "loop is not handled by vectorizer");
+ return false;
+ }
- PtrRtCheck.Need = NeedRTCheck;
+ continue;
+ }
- bool CanVecMem = true;
- if (Accesses.isDependencyCheckNeeded()) {
- DEBUG(dbgs() << "LV: Checking memory dependencies\n");
- CanVecMem = DepChecker.areDepsSafe(
- DependentAccesses, Accesses.getDependenciesToCheck(), Strides);
- MaxSafeDepDistBytes = DepChecker.getMaxSafeDepDistBytes();
+ if (RecurrenceDescriptor::isReductionPHI(Phi, TheLoop,
+ Reductions[Phi])) {
+ if (Reductions[Phi].hasUnsafeAlgebra())
+ Requirements->addUnsafeAlgebraInst(
+ Reductions[Phi].getUnsafeAlgebraInst());
+ AllowedExit.insert(Reductions[Phi].getLoopExitInstr());
+ continue;
+ }
- if (!CanVecMem && DepChecker.shouldRetryWithRuntimeCheck()) {
- DEBUG(dbgs() << "LV: Retrying with memory checks\n");
- NeedRTCheck = true;
+ emitAnalysis(VectorizationReport(it) <<
+ "value that could not be identified as "
+ "reduction is used outside the loop");
+ DEBUG(dbgs() << "LV: Found an unidentified PHI."<< *Phi <<"\n");
+ return false;
+ }// end of PHI handling
- // Clear the dependency checks. We assume they are not needed.
- Accesses.resetDepChecks();
+ // We handle calls that:
+ // * Are debug info intrinsics.
+ // * Have a mapping to an IR intrinsic.
+ // * Have a vector version available.
+ CallInst *CI = dyn_cast<CallInst>(it);
+ if (CI && !getIntrinsicIDForCall(CI, TLI) && !isa<DbgInfoIntrinsic>(CI) &&
+ !(CI->getCalledFunction() && TLI &&
+ TLI->isFunctionVectorizable(CI->getCalledFunction()->getName()))) {
+ emitAnalysis(VectorizationReport(it) <<
+ "call instruction cannot be vectorized");
+ DEBUG(dbgs() << "LV: Found a non-intrinsic, non-libfunc callsite.\n");
+ return false;
+ }
- PtrRtCheck.reset();
- PtrRtCheck.Need = true;
+ // Intrinsics such as powi,cttz and ctlz are legal to vectorize if the
+ // second argument is the same (i.e. loop invariant)
+ if (CI &&
+ hasVectorInstrinsicScalarOpd(getIntrinsicIDForCall(CI, TLI), 1)) {
+ if (!SE->isLoopInvariant(SE->getSCEV(CI->getOperand(1)), TheLoop)) {
+ emitAnalysis(VectorizationReport(it)
+ << "intrinsic instruction cannot be vectorized");
+ DEBUG(dbgs() << "LV: Found unvectorizable intrinsic " << *CI << "\n");
+ return false;
+ }
+ }
- CanDoRT = Accesses.canCheckPtrAtRT(PtrRtCheck, NumComparisons, SE,
- TheLoop, Strides, true);
- // Check that we did not collect too many pointers or found an unsizeable
- // pointer.
- if (!CanDoRT || NumComparisons > VectParams.RuntimeMemoryCheckThreshold) {
- if (!CanDoRT && NumComparisons > 0)
- emitAnalysis(VectorizationReport()
- << "cannot check memory dependencies at runtime");
- else
- emitAnalysis(VectorizationReport()
- << NumComparisons << " exceeds limit of "
- << VectParams.RuntimeMemoryCheckThreshold
- << " dependent memory operations checked at runtime");
- DEBUG(dbgs() << "LV: Can't vectorize with memory checks\n");
- PtrRtCheck.reset();
+ // Check that the instruction return type is vectorizable.
+ // Also, we can't vectorize extractelement instructions.
+ if ((!VectorType::isValidElementType(it->getType()) &&
+ !it->getType()->isVoidTy()) || isa<ExtractElementInst>(it)) {
+ emitAnalysis(VectorizationReport(it)
+ << "instruction return type cannot be vectorized");
+ DEBUG(dbgs() << "LV: Found unvectorizable type.\n");
return false;
}
- CanVecMem = true;
- }
- }
-
- if (!CanVecMem)
- emitAnalysis(VectorizationReport() <<
- "unsafe dependent memory operations in loop");
+ // Check that the stored type is vectorizable.
+ if (StoreInst *ST = dyn_cast<StoreInst>(it)) {
+ Type *T = ST->getValueOperand()->getType();
+ if (!VectorType::isValidElementType(T)) {
+ emitAnalysis(VectorizationReport(ST) <<
+ "store instruction cannot be vectorized");
+ return false;
+ }
+ if (EnableMemAccessVersioning)
+ collectStridedAccess(ST);
+ }
- DEBUG(dbgs() << "LV: We" << (NeedRTCheck ? "" : " don't") <<
- " need a runtime memory check.\n");
+ if (EnableMemAccessVersioning)
+ if (LoadInst *LI = dyn_cast<LoadInst>(it))
+ collectStridedAccess(LI);
- return CanVecMem;
-}
+ // Reduction instructions are allowed to have exit users.
+ // All other instructions must not have external users.
+ if (hasOutsideLoopUser(TheLoop, it, AllowedExit)) {
+ emitAnalysis(VectorizationReport(it) <<
+ "value cannot be used outside the loop");
+ return false;
+ }
-bool LoopVectorizationLegality::canVectorizeMemory() {
- return LAA.canVectorizeMemory(Strides);
-}
+ } // next instr.
-static bool hasMultipleUsesOf(Instruction *I,
- SmallPtrSetImpl<Instruction *> &Insts) {
- unsigned NumUses = 0;
- for(User::op_iterator Use = I->op_begin(), E = I->op_end(); Use != E; ++Use) {
- if (Insts.count(dyn_cast<Instruction>(*Use)))
- ++NumUses;
- if (NumUses > 1)
- return true;
}
- return false;
-}
-
-static bool areAllUsesIn(Instruction *I, SmallPtrSetImpl<Instruction *> &Set) {
- for(User::op_iterator Use = I->op_begin(), E = I->op_end(); Use != E; ++Use)
- if (!Set.count(dyn_cast<Instruction>(*Use)))
+ if (!Induction) {
+ DEBUG(dbgs() << "LV: Did not find one integer induction var.\n");
+ if (Inductions.empty()) {
+ emitAnalysis(VectorizationReport()
+ << "loop induction variable could not be identified");
return false;
+ }
+ }
+
return true;
}
-bool LoopVectorizationLegality::AddReductionVar(PHINode *Phi,
- ReductionKind Kind) {
- if (Phi->getNumIncomingValues() != 2)
- return false;
-
- // Reduction variables are only found in the loop header block.
- if (Phi->getParent() != TheLoop->getHeader())
- return false;
-
- // Obtain the reduction start value from the value that comes from the loop
- // preheader.
- Value *RdxStart = Phi->getIncomingValueForBlock(TheLoop->getLoopPreheader());
-
- // ExitInstruction is the single value which is used outside the loop.
- // We only allow for a single reduction value to be used outside the loop.
- // This includes users of the reduction, variables (which form a cycle
- // which ends in the phi node).
- Instruction *ExitInstruction = nullptr;
- // Indicates that we found a reduction operation in our scan.
- bool FoundReduxOp = false;
-
- // We start with the PHI node and scan for all of the users of this
- // instruction. All users must be instructions that can be used as reduction
- // variables (such as ADD). We must have a single out-of-block user. The cycle
- // must include the original PHI.
- bool FoundStartPHI = false;
-
- // To recognize min/max patterns formed by a icmp select sequence, we store
- // the number of instruction we saw from the recognized min/max pattern,
- // to make sure we only see exactly the two instructions.
- unsigned NumCmpSelectPatternInst = 0;
- ReductionInstDesc ReduxDesc(false, nullptr);
-
- SmallPtrSet<Instruction *, 8> VisitedInsts;
- SmallVector<Instruction *, 8> Worklist;
- Worklist.push_back(Phi);
- VisitedInsts.insert(Phi);
-
- // A value in the reduction can be used:
- // - By the reduction:
- // - Reduction operation:
- // - One use of reduction value (safe).
- // - Multiple use of reduction value (not safe).
- // - PHI:
- // - All uses of the PHI must be the reduction (safe).
- // - Otherwise, not safe.
- // - By one instruction outside of the loop (safe).
- // - By further instructions outside of the loop (not safe).
- // - By an instruction that is not part of the reduction (not safe).
- // This is either:
- // * An instruction type other than PHI or the reduction operation.
- // * A PHI in the header other than the initial PHI.
- while (!Worklist.empty()) {
- Instruction *Cur = Worklist.back();
- Worklist.pop_back();
-
- // No Users.
- // If the instruction has no users then this is a broken chain and can't be
- // a reduction variable.
- if (Cur->use_empty())
- return false;
-
- bool IsAPhi = isa<PHINode>(Cur);
-
- // A header PHI use other than the original PHI.
- if (Cur != Phi && IsAPhi && Cur->getParent() == Phi->getParent())
- return false;
-
- // Reductions of instructions such as Div, and Sub is only possible if the
- // LHS is the reduction variable.
- if (!Cur->isCommutative() && !IsAPhi && !isa<SelectInst>(Cur) &&
- !isa<ICmpInst>(Cur) && !isa<FCmpInst>(Cur) &&
- !VisitedInsts.count(dyn_cast<Instruction>(Cur->getOperand(0))))
- return false;
+void LoopVectorizationLegality::collectStridedAccess(Value *MemAccess) {
+ Value *Ptr = nullptr;
+ if (LoadInst *LI = dyn_cast<LoadInst>(MemAccess))
+ Ptr = LI->getPointerOperand();
+ else if (StoreInst *SI = dyn_cast<StoreInst>(MemAccess))
+ Ptr = SI->getPointerOperand();
+ else
+ return;
- // Any reduction instruction must be of one of the allowed kinds.
- ReduxDesc = isReductionInstr(Cur, Kind, ReduxDesc);
- if (!ReduxDesc.IsReduction)
- return false;
+ Value *Stride = getStrideFromPointer(Ptr, SE, TheLoop);
+ if (!Stride)
+ return;
- // A reduction operation must only have one use of the reduction value.
- if (!IsAPhi && Kind != RK_IntegerMinMax && Kind != RK_FloatMinMax &&
- hasMultipleUsesOf(Cur, VisitedInsts))
- return false;
+ DEBUG(dbgs() << "LV: Found a strided access that we can version");
+ DEBUG(dbgs() << " Ptr: " << *Ptr << " Stride: " << *Stride << "\n");
+ Strides[Ptr] = Stride;
+ StrideSet.insert(Stride);
+}
- // All inputs to a PHI node must be a reduction value.
- if(IsAPhi && Cur != Phi && !areAllUsesIn(Cur, VisitedInsts))
- return false;
+void LoopVectorizationLegality::collectLoopUniforms() {
+ // We now know that the loop is vectorizable!
+ // Collect variables that will remain uniform after vectorization.
+ std::vector<Value*> Worklist;
+ BasicBlock *Latch = TheLoop->getLoopLatch();
- if (Kind == RK_IntegerMinMax && (isa<ICmpInst>(Cur) ||
- isa<SelectInst>(Cur)))
- ++NumCmpSelectPatternInst;
- if (Kind == RK_FloatMinMax && (isa<FCmpInst>(Cur) ||
- isa<SelectInst>(Cur)))
- ++NumCmpSelectPatternInst;
-
- // Check whether we found a reduction operator.
- FoundReduxOp |= !IsAPhi;
-
- // Process users of current instruction. Push non-PHI nodes after PHI nodes
- // onto the stack. This way we are going to have seen all inputs to PHI
- // nodes once we get to them.
- SmallVector<Instruction *, 8> NonPHIs;
- SmallVector<Instruction *, 8> PHIs;
- for (User *U : Cur->users()) {
- Instruction *UI = cast<Instruction>(U);
+ // Start with the conditional branch and walk up the block.
+ Worklist.push_back(Latch->getTerminator()->getOperand(0));
- // Check if we found the exit user.
- BasicBlock *Parent = UI->getParent();
- if (!TheLoop->contains(Parent)) {
- // Exit if you find multiple outside users or if the header phi node is
- // being used. In this case the user uses the value of the previous
- // iteration, in which case we would loose "VF-1" iterations of the
- // reduction operation if we vectorize.
- if (ExitInstruction != nullptr || Cur == Phi)
- return false;
+ // Also add all consecutive pointer values; these values will be uniform
+ // after vectorization (and subsequent cleanup) and, until revectorization is
+ // supported, all dependencies must also be uniform.
+ for (Loop::block_iterator B = TheLoop->block_begin(),
+ BE = TheLoop->block_end(); B != BE; ++B)
+ for (BasicBlock::iterator I = (*B)->begin(), IE = (*B)->end();
+ I != IE; ++I)
+ if (I->getType()->isPointerTy() && isConsecutivePtr(I))
+ Worklist.insert(Worklist.end(), I->op_begin(), I->op_end());
- // The instruction used by an outside user must be the last instruction
- // before we feed back to the reduction phi. Otherwise, we loose VF-1
- // operations on the value.
- if (std::find(Phi->op_begin(), Phi->op_end(), Cur) == Phi->op_end())
- return false;
+ while (!Worklist.empty()) {
+ Instruction *I = dyn_cast<Instruction>(Worklist.back());
+ Worklist.pop_back();
- ExitInstruction = Cur;
- continue;
- }
+ // Look at instructions inside this loop.
+ // Stop when reaching PHI nodes.
+ // TODO: we need to follow values all over the loop, not only in this block.
+ if (!I || !TheLoop->contains(I) || isa<PHINode>(I))
+ continue;
- // Process instructions only once (termination). Each reduction cycle
- // value must only be used once, except by phi nodes and min/max
- // reductions which are represented as a cmp followed by a select.
- ReductionInstDesc IgnoredVal(false, nullptr);
- if (VisitedInsts.insert(UI).second) {
- if (isa<PHINode>(UI))
- PHIs.push_back(UI);
- else
- NonPHIs.push_back(UI);
- } else if (!isa<PHINode>(UI) &&
- ((!isa<FCmpInst>(UI) &&
- !isa<ICmpInst>(UI) &&
- !isa<SelectInst>(UI)) ||
- !isMinMaxSelectCmpPattern(UI, IgnoredVal).IsReduction))
- return false;
+ // This is a known uniform.
+ Uniforms.insert(I);
- // Remember that we completed the cycle.
- if (UI == Phi)
- FoundStartPHI = true;
- }
- Worklist.append(PHIs.begin(), PHIs.end());
- Worklist.append(NonPHIs.begin(), NonPHIs.end());
+ // Insert all operands.
+ Worklist.insert(Worklist.end(), I->op_begin(), I->op_end());
}
+}
- // This means we have seen one but not the other instruction of the
- // pattern or more than just a select and cmp.
- if ((Kind == RK_IntegerMinMax || Kind == RK_FloatMinMax) &&
- NumCmpSelectPatternInst != 2)
+bool LoopVectorizationLegality::canVectorizeMemory() {
+ LAI = &LAA->getInfo(TheLoop, Strides);
+ auto &OptionalReport = LAI->getReport();
+ if (OptionalReport)
+ emitAnalysis(VectorizationReport(*OptionalReport));
+ if (!LAI->canVectorizeMemory())
return false;
- if (!FoundStartPHI || !FoundReduxOp || !ExitInstruction)
+ if (LAI->hasStoreToLoopInvariantAddress()) {
+ emitAnalysis(
+ VectorizationReport()
+ << "write to a loop invariant address could not be vectorized");
+ DEBUG(dbgs() << "LV: We don't allow storing to uniform addresses\n");
return false;
+ }
- // We found a reduction var if we have reached the original phi node and we
- // only have a single instruction with out-of-loop users.
-
- // This instruction is allowed to have out-of-loop users.
- AllowedExit.insert(ExitInstruction);
-
- // Save the description of this reduction variable.
- ReductionDescriptor RD(RdxStart, ExitInstruction, Kind,
- ReduxDesc.MinMaxKind);
- Reductions[Phi] = RD;
- // We've ended the cycle. This is a reduction variable if we have an
- // outside user and it has a binary op.
+ Requirements->addRuntimePointerChecks(LAI->getNumRuntimePointerChecks());
return true;
}
-/// Returns true if the instruction is a Select(ICmp(X, Y), X, Y) instruction
-/// pattern corresponding to a min(X, Y) or max(X, Y).
-LoopVectorizationLegality::ReductionInstDesc
-LoopVectorizationLegality::isMinMaxSelectCmpPattern(Instruction *I,
- ReductionInstDesc &Prev) {
-
- assert((isa<ICmpInst>(I) || isa<FCmpInst>(I) || isa<SelectInst>(I)) &&
- "Expect a select instruction");
- Instruction *Cmp = nullptr;
- SelectInst *Select = nullptr;
-
- // We must handle the select(cmp()) as a single instruction. Advance to the
- // select.
- if ((Cmp = dyn_cast<ICmpInst>(I)) || (Cmp = dyn_cast<FCmpInst>(I))) {
- if (!Cmp->hasOneUse() || !(Select = dyn_cast<SelectInst>(*I->user_begin())))
- return ReductionInstDesc(false, I);
- return ReductionInstDesc(Select, Prev.MinMaxKind);
- }
-
- // Only handle single use cases for now.
- if (!(Select = dyn_cast<SelectInst>(I)))
- return ReductionInstDesc(false, I);
- if (!(Cmp = dyn_cast<ICmpInst>(I->getOperand(0))) &&
- !(Cmp = dyn_cast<FCmpInst>(I->getOperand(0))))
- return ReductionInstDesc(false, I);
- if (!Cmp->hasOneUse())
- return ReductionInstDesc(false, I);
-
- Value *CmpLeft;
- Value *CmpRight;
-
- // Look for a min/max pattern.
- if (m_UMin(m_Value(CmpLeft), m_Value(CmpRight)).match(Select))
- return ReductionInstDesc(Select, MRK_UIntMin);
- else if (m_UMax(m_Value(CmpLeft), m_Value(CmpRight)).match(Select))
- return ReductionInstDesc(Select, MRK_UIntMax);
- else if (m_SMax(m_Value(CmpLeft), m_Value(CmpRight)).match(Select))
- return ReductionInstDesc(Select, MRK_SIntMax);
- else if (m_SMin(m_Value(CmpLeft), m_Value(CmpRight)).match(Select))
- return ReductionInstDesc(Select, MRK_SIntMin);
- else if (m_OrdFMin(m_Value(CmpLeft), m_Value(CmpRight)).match(Select))
- return ReductionInstDesc(Select, MRK_FloatMin);
- else if (m_OrdFMax(m_Value(CmpLeft), m_Value(CmpRight)).match(Select))
- return ReductionInstDesc(Select, MRK_FloatMax);
- else if (m_UnordFMin(m_Value(CmpLeft), m_Value(CmpRight)).match(Select))
- return ReductionInstDesc(Select, MRK_FloatMin);
- else if (m_UnordFMax(m_Value(CmpLeft), m_Value(CmpRight)).match(Select))
- return ReductionInstDesc(Select, MRK_FloatMax);
-
- return ReductionInstDesc(false, I);
-}
-
-LoopVectorizationLegality::ReductionInstDesc
-LoopVectorizationLegality::isReductionInstr(Instruction *I,
- ReductionKind Kind,
- ReductionInstDesc &Prev) {
- bool FP = I->getType()->isFloatingPointTy();
- bool FastMath = FP && I->hasUnsafeAlgebra();
- switch (I->getOpcode()) {
- default:
- return ReductionInstDesc(false, I);
- case Instruction::PHI:
- if (FP && (Kind != RK_FloatMult && Kind != RK_FloatAdd &&
- Kind != RK_FloatMinMax))
- return ReductionInstDesc(false, I);
- return ReductionInstDesc(I, Prev.MinMaxKind);
- case Instruction::Sub:
- case Instruction::Add:
- return ReductionInstDesc(Kind == RK_IntegerAdd, I);
- case Instruction::Mul:
- return ReductionInstDesc(Kind == RK_IntegerMult, I);
- case Instruction::And:
- return ReductionInstDesc(Kind == RK_IntegerAnd, I);
- case Instruction::Or:
- return ReductionInstDesc(Kind == RK_IntegerOr, I);
- case Instruction::Xor:
- return ReductionInstDesc(Kind == RK_IntegerXor, I);
- case Instruction::FMul:
- return ReductionInstDesc(Kind == RK_FloatMult && FastMath, I);
- case Instruction::FSub:
- case Instruction::FAdd:
- return ReductionInstDesc(Kind == RK_FloatAdd && FastMath, I);
- case Instruction::FCmp:
- case Instruction::ICmp:
- case Instruction::Select:
- if (Kind != RK_IntegerMinMax &&
- (!HasFunNoNaNAttr || Kind != RK_FloatMinMax))
- return ReductionInstDesc(false, I);
- return isMinMaxSelectCmpPattern(I, Prev);
- }
-}
-
LoopVectorizationLegality::InductionKind
LoopVectorizationLegality::isInductionVariable(PHINode *Phi,
ConstantInt *&StepValue) {
- Type *PhiTy = Phi->getType();
- // We only handle integer and pointer inductions variables.
- if (!PhiTy->isIntegerTy() && !PhiTy->isPointerTy())
- return IK_NoInduction;
-
- // Check that the PHI is consecutive.
- const SCEV *PhiScev = SE->getSCEV(Phi);
- const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(PhiScev);
- if (!AR) {
- DEBUG(dbgs() << "LV: PHI is not a poly recurrence.\n");
- return IK_NoInduction;
- }
-
- 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)
+ if (!isInductionPHI(Phi, SE, StepValue))
return IK_NoInduction;
- ConstantInt *CV = C->getValue();
- if (PhiTy->isIntegerTy()) {
- StepValue = CV;
+ Type *PhiTy = Phi->getType();
+ // Found an Integer induction variable.
+ if (PhiTy->isIntegerTy())
return IK_IntInduction;
- }
-
- assert(PhiTy->isPointerTy() && "The PHI must be a pointer");
- Type *PointerElementType = PhiTy->getPointerElementType();
- // The pointer stride cannot be determined if the pointer element type is not
- // sized.
- if (!PointerElementType->isSized())
- return IK_NoInduction;
-
- int64_t Size = static_cast<int64_t>(DL->getTypeAllocSize(PointerElementType));
- int64_t CVSize = CV->getSExtValue();
- if (CVSize % Size)
- return IK_NoInduction;
- StepValue = ConstantInt::getSigned(CV->getType(), CVSize / Size);
+ // Found an Pointer induction variable.
return IK_PtrInduction;
}
return Inductions.count(PN);
}
-bool LoopAccessAnalysis::blockNeedsPredication(BasicBlock *BB) {
- assert(TheLoop->contains(BB) && "Unknown block used");
-
- // Blocks that do not dominate the latch need predication.
- BasicBlock* Latch = TheLoop->getLoopLatch();
- return !DT->dominates(BB, Latch);
-}
-
bool LoopVectorizationLegality::blockNeedsPredication(BasicBlock *BB) {
- return LAA.blockNeedsPredication(BB);
+ return LoopAccessInfo::blockNeedsPredication(BB, TheLoop, DT);
}
bool LoopVectorizationLegality::blockCanBePredicated(BasicBlock *BB,
return true;
}
+void InterleavedAccessInfo::collectConstStridedAccesses(
+ MapVector<Instruction *, StrideDescriptor> &StrideAccesses,
+ const ValueToValueMap &Strides) {
+ // Holds load/store instructions in program order.
+ SmallVector<Instruction *, 16> AccessList;
+
+ for (auto *BB : TheLoop->getBlocks()) {
+ bool IsPred = LoopAccessInfo::blockNeedsPredication(BB, TheLoop, DT);
+
+ for (auto &I : *BB) {
+ if (!isa<LoadInst>(&I) && !isa<StoreInst>(&I))
+ continue;
+ // FIXME: Currently we can't handle mixed accesses and predicated accesses
+ if (IsPred)
+ return;
+
+ AccessList.push_back(&I);
+ }
+ }
+
+ if (AccessList.empty())
+ return;
+
+ auto &DL = TheLoop->getHeader()->getModule()->getDataLayout();
+ for (auto I : AccessList) {
+ LoadInst *LI = dyn_cast<LoadInst>(I);
+ StoreInst *SI = dyn_cast<StoreInst>(I);
+
+ Value *Ptr = LI ? LI->getPointerOperand() : SI->getPointerOperand();
+ int Stride = isStridedPtr(SE, Ptr, TheLoop, Strides);
+
+ // The factor of the corresponding interleave group.
+ unsigned Factor = std::abs(Stride);
+
+ // Ignore the access if the factor is too small or too large.
+ if (Factor < 2 || Factor > MaxInterleaveGroupFactor)
+ continue;
+
+ const SCEV *Scev = replaceSymbolicStrideSCEV(SE, Strides, Ptr);
+ PointerType *PtrTy = dyn_cast<PointerType>(Ptr->getType());
+ unsigned Size = DL.getTypeAllocSize(PtrTy->getElementType());
+
+ // An alignment of 0 means target ABI alignment.
+ unsigned Align = LI ? LI->getAlignment() : SI->getAlignment();
+ if (!Align)
+ Align = DL.getABITypeAlignment(PtrTy->getElementType());
+
+ StrideAccesses[I] = StrideDescriptor(Stride, Scev, Size, Align);
+ }
+}
+
+// Analyze interleaved accesses and collect them into interleave groups.
+//
+// Notice that the vectorization on interleaved groups will change instruction
+// orders and may break dependences. But the memory dependence check guarantees
+// that there is no overlap between two pointers of different strides, element
+// sizes or underlying bases.
+//
+// For pointers sharing the same stride, element size and underlying base, no
+// need to worry about Read-After-Write dependences and Write-After-Read
+// dependences.
+//
+// E.g. The RAW dependence: A[i] = a;
+// b = A[i];
+// This won't exist as it is a store-load forwarding conflict, which has
+// already been checked and forbidden in the dependence check.
+//
+// E.g. The WAR dependence: a = A[i]; // (1)
+// A[i] = b; // (2)
+// The store group of (2) is always inserted at or below (2), and the load group
+// of (1) is always inserted at or above (1). The dependence is safe.
+void InterleavedAccessInfo::analyzeInterleaving(
+ const ValueToValueMap &Strides) {
+ DEBUG(dbgs() << "LV: Analyzing interleaved accesses...\n");
+
+ // Holds all the stride accesses.
+ MapVector<Instruction *, StrideDescriptor> StrideAccesses;
+ collectConstStridedAccesses(StrideAccesses, Strides);
+
+ if (StrideAccesses.empty())
+ return;
+
+ // Holds all interleaved store groups temporarily.
+ SmallSetVector<InterleaveGroup *, 4> StoreGroups;
+
+ // Search the load-load/write-write pair B-A in bottom-up order and try to
+ // insert B into the interleave group of A according to 3 rules:
+ // 1. A and B have the same stride.
+ // 2. A and B have the same memory object size.
+ // 3. B belongs to the group according to the distance.
+ //
+ // The bottom-up order can avoid breaking the Write-After-Write dependences
+ // between two pointers of the same base.
+ // E.g. A[i] = a; (1)
+ // A[i] = b; (2)
+ // A[i+1] = c (3)
+ // We form the group (2)+(3) in front, so (1) has to form groups with accesses
+ // above (1), which guarantees that (1) is always above (2).
+ for (auto I = StrideAccesses.rbegin(), E = StrideAccesses.rend(); I != E;
+ ++I) {
+ Instruction *A = I->first;
+ StrideDescriptor DesA = I->second;
+
+ InterleaveGroup *Group = getInterleaveGroup(A);
+ if (!Group) {
+ DEBUG(dbgs() << "LV: Creating an interleave group with:" << *A << '\n');
+ Group = createInterleaveGroup(A, DesA.Stride, DesA.Align);
+ }
+
+ if (A->mayWriteToMemory())
+ StoreGroups.insert(Group);
+
+ for (auto II = std::next(I); II != E; ++II) {
+ Instruction *B = II->first;
+ StrideDescriptor DesB = II->second;
+
+ // Ignore if B is already in a group or B is a different memory operation.
+ if (isInterleaved(B) || A->mayReadFromMemory() != B->mayReadFromMemory())
+ continue;
+
+ // Check the rule 1 and 2.
+ if (DesB.Stride != DesA.Stride || DesB.Size != DesA.Size)
+ continue;
+
+ // Calculate the distance and prepare for the rule 3.
+ const SCEVConstant *DistToA =
+ dyn_cast<SCEVConstant>(SE->getMinusSCEV(DesB.Scev, DesA.Scev));
+ if (!DistToA)
+ continue;
+
+ int DistanceToA = DistToA->getValue()->getValue().getSExtValue();
+
+ // Skip if the distance is not multiple of size as they are not in the
+ // same group.
+ if (DistanceToA % static_cast<int>(DesA.Size))
+ continue;
+
+ // The index of B is the index of A plus the related index to A.
+ int IndexB =
+ Group->getIndex(A) + DistanceToA / static_cast<int>(DesA.Size);
+
+ // Try to insert B into the group.
+ if (Group->insertMember(B, IndexB, DesB.Align)) {
+ DEBUG(dbgs() << "LV: Inserted:" << *B << '\n'
+ << " into the interleave group with" << *A << '\n');
+ InterleaveGroupMap[B] = Group;
+
+ // Set the first load in program order as the insert position.
+ if (B->mayReadFromMemory())
+ Group->setInsertPos(B);
+ }
+ } // Iteration on instruction B
+ } // Iteration on instruction A
+
+ // Remove interleaved store groups with gaps.
+ for (InterleaveGroup *Group : StoreGroups)
+ if (Group->getNumMembers() != Group->getFactor())
+ releaseGroup(Group);
+}
+
LoopVectorizationCostModel::VectorizationFactor
LoopVectorizationCostModel::selectVectorizationFactor(bool OptForSize) {
// Width 1 means no vectorize
VectorizationFactor Factor = { 1U, 0U };
- if (OptForSize && Legal->getRuntimePointerCheck()->Need) {
+ if (OptForSize && Legal->getRuntimePointerChecking()->Need) {
emitAnalysis(VectorizationReport() <<
"runtime pointer checks needed. Enable vectorization of this "
"loop with '#pragma clang loop vectorize(enable)' when "
- "compiling with -Os");
- DEBUG(dbgs() << "LV: Aborting. Runtime ptr check is required in Os.\n");
+ "compiling with -Os/-Oz");
+ DEBUG(dbgs() <<
+ "LV: Aborting. Runtime ptr check is required with -Os/-Oz.\n");
return Factor;
}
emitAnalysis
(VectorizationReport() <<
"unable to calculate the loop count due to complex control flow");
- DEBUG(dbgs() << "LV: Aborting. A tail loop is required in Os.\n");
+ DEBUG(dbgs() << "LV: Aborting. A tail loop is required with -Os/-Oz.\n");
return Factor;
}
if (VF == 0)
VF = MaxVectorSize;
-
- // If the trip count that we found modulo the vectorization factor is not
- // zero then we require a tail.
- if (VF < 2) {
+ else {
+ // If the trip count that we found modulo the vectorization factor is not
+ // zero then we require a tail.
emitAnalysis(VectorizationReport() <<
"cannot optimize for size and vectorize at the "
"same time. Enable vectorization of this loop "
"with '#pragma clang loop vectorize(enable)' "
- "when compiling with -Os");
- DEBUG(dbgs() << "LV: Aborting. A tail loop is required in Os.\n");
+ "when compiling with -Os/-Oz");
+ DEBUG(dbgs() << "LV: Aborting. A tail loop is required with -Os/-Oz.\n");
return Factor;
}
}
unsigned LoopVectorizationCostModel::getWidestType() {
unsigned MaxWidth = 8;
+ const DataLayout &DL = TheFunction->getParent()->getDataLayout();
// For each block.
for (Loop::block_iterator bb = TheLoop->block_begin(),
continue;
MaxWidth = std::max(MaxWidth,
- (unsigned)DL->getTypeSizeInBits(T->getScalarType()));
+ (unsigned)DL.getTypeSizeInBits(T->getScalarType()));
}
}
return MaxWidth;
}
-unsigned
-LoopVectorizationCostModel::selectUnrollFactor(bool OptForSize,
- unsigned VF,
- unsigned LoopCost) {
+unsigned LoopVectorizationCostModel::selectInterleaveCount(bool OptForSize,
+ unsigned VF,
+ unsigned LoopCost) {
- // -- The unroll heuristics --
- // We unroll the loop in order to expose ILP and reduce the loop overhead.
+ // -- The interleave heuristics --
+ // We interleave the loop in order to expose ILP and reduce the loop overhead.
// There are many micro-architectural considerations that we can't predict
// at this level. For example, frontend pressure (on decode or fetch) due to
// code size, or the number and capabilities of the execution ports.
//
- // We use the following heuristics to select the unroll factor:
- // 1. If the code has reductions, then we unroll in order to break the cross
+ // We use the following heuristics to select the interleave count:
+ // 1. If the code has reductions, then we interleave to break the cross
// iteration dependency.
- // 2. If the loop is really small, then we unroll in order to reduce the loop
+ // 2. If the loop is really small, then we interleave to reduce the loop
// overhead.
- // 3. We don't unroll if we think that we will spill registers to memory due
- // to the increased register pressure.
-
- // Use the user preference, unless 'auto' is selected.
- int UserUF = Hints->getInterleave();
- if (UserUF != 0)
- return UserUF;
+ // 3. We don't interleave if we think that we will spill registers to memory
+ // due to the increased register pressure.
- // When we optimize for size, we don't unroll.
+ // When we optimize for size, we don't interleave.
if (OptForSize)
return 1;
- // We used the distance for the unroll factor.
+ // We used the distance for the interleave count.
if (Legal->getMaxSafeDepDistBytes() != -1U)
return 1;
- // Do not unroll loops with a relatively small trip count.
+ // Do not interleave loops with a relatively small trip count.
unsigned TC = SE->getSmallConstantTripCount(TheLoop);
- if (TC > 1 && TC < TinyTripCountUnrollThreshold)
+ if (TC > 1 && TC < TinyTripCountInterleaveThreshold)
return 1;
unsigned TargetNumRegisters = TTI.getNumberOfRegisters(VF > 1);
R.MaxLocalUsers = std::max(R.MaxLocalUsers, 1U);
R.NumInstructions = std::max(R.NumInstructions, 1U);
- // We calculate the unroll factor using the following formula.
+ // We calculate the interleave count using the following formula.
// Subtract the number of loop invariants from the number of available
- // registers. These registers are used by all of the unrolled instances.
+ // registers. These registers are used by all of the interleaved instances.
// Next, divide the remaining registers by the number of registers that is
// required by the loop, in order to estimate how many parallel instances
// fit without causing spills. All of this is rounded down if necessary to be
- // a power of two. We want power of two unroll factors to simplify any
+ // a power of two. We want power of two interleave count to simplify any
// addressing operations or alignment considerations.
- unsigned UF = PowerOf2Floor((TargetNumRegisters - R.LoopInvariantRegs) /
+ unsigned IC = PowerOf2Floor((TargetNumRegisters - R.LoopInvariantRegs) /
R.MaxLocalUsers);
- // Don't count the induction variable as unrolled.
+ // Don't count the induction variable as interleaved.
if (EnableIndVarRegisterHeur)
- UF = PowerOf2Floor((TargetNumRegisters - R.LoopInvariantRegs - 1) /
+ IC = PowerOf2Floor((TargetNumRegisters - R.LoopInvariantRegs - 1) /
std::max(1U, (R.MaxLocalUsers - 1)));
- // Clamp the unroll factor ranges to reasonable factors.
- unsigned MaxInterleaveSize = TTI.getMaxInterleaveFactor();
+ // Clamp the interleave ranges to reasonable counts.
+ unsigned MaxInterleaveCount = TTI.getMaxInterleaveFactor(VF);
- // Check if the user has overridden the unroll max.
+ // Check if the user has overridden the max.
if (VF == 1) {
if (ForceTargetMaxScalarInterleaveFactor.getNumOccurrences() > 0)
- MaxInterleaveSize = ForceTargetMaxScalarInterleaveFactor;
+ MaxInterleaveCount = ForceTargetMaxScalarInterleaveFactor;
} else {
if (ForceTargetMaxVectorInterleaveFactor.getNumOccurrences() > 0)
- MaxInterleaveSize = ForceTargetMaxVectorInterleaveFactor;
+ MaxInterleaveCount = ForceTargetMaxVectorInterleaveFactor;
}
// If we did not calculate the cost for VF (because the user selected the VF)
if (LoopCost == 0)
LoopCost = expectedCost(VF);
- // Clamp the calculated UF to be between the 1 and the max unroll factor
+ // Clamp the calculated IC to be between the 1 and the max interleave count
// that the target allows.
- if (UF > MaxInterleaveSize)
- UF = MaxInterleaveSize;
- else if (UF < 1)
- UF = 1;
+ if (IC > MaxInterleaveCount)
+ IC = MaxInterleaveCount;
+ else if (IC < 1)
+ IC = 1;
- // Unroll if we vectorized this loop and there is a reduction that could
- // benefit from unrolling.
+ // Interleave if we vectorized this loop and there is a reduction that could
+ // benefit from interleaving.
if (VF > 1 && Legal->getReductionVars()->size()) {
- DEBUG(dbgs() << "LV: Unrolling because of reductions.\n");
- return UF;
+ DEBUG(dbgs() << "LV: Interleaving because of reductions.\n");
+ return IC;
}
// Note that if we've already vectorized the loop we will have done the
- // runtime check and so unrolling won't require further checks.
- bool UnrollingRequiresRuntimePointerCheck =
- (VF == 1 && Legal->getRuntimePointerCheck()->Need);
+ // runtime check and so interleaving won't require further checks.
+ bool InterleavingRequiresRuntimePointerCheck =
+ (VF == 1 && Legal->getRuntimePointerChecking()->Need);
- // We want to unroll small loops in order to reduce the loop overhead and
+ // We want to interleave small loops in order to reduce the loop overhead and
// potentially expose ILP opportunities.
DEBUG(dbgs() << "LV: Loop cost is " << LoopCost << '\n');
- if (!UnrollingRequiresRuntimePointerCheck &&
- LoopCost < SmallLoopCost) {
+ if (!InterleavingRequiresRuntimePointerCheck && LoopCost < SmallLoopCost) {
// We assume that the cost overhead is 1 and we use the cost model
- // to estimate the cost of the loop and unroll until the cost of the
+ // to estimate the cost of the loop and interleave until the cost of the
// loop overhead is about 5% of the cost of the loop.
- unsigned SmallUF = std::min(UF, (unsigned)PowerOf2Floor(SmallLoopCost / LoopCost));
+ unsigned SmallIC =
+ std::min(IC, (unsigned)PowerOf2Floor(SmallLoopCost / LoopCost));
- // Unroll until store/load ports (estimated by max unroll factor) are
+ // Interleave until store/load ports (estimated by max interleave count) are
// saturated.
unsigned NumStores = Legal->getNumStores();
unsigned NumLoads = Legal->getNumLoads();
- unsigned StoresUF = UF / (NumStores ? NumStores : 1);
- unsigned LoadsUF = UF / (NumLoads ? NumLoads : 1);
+ unsigned StoresIC = IC / (NumStores ? NumStores : 1);
+ unsigned LoadsIC = IC / (NumLoads ? NumLoads : 1);
// If we have a scalar reduction (vector reductions are already dealt with
// by this point), we can increase the critical path length if the loop
- // we're unrolling is inside another loop. Limit, by default to 2, so the
+ // we're interleaving is inside another loop. Limit, by default to 2, so the
// critical path only gets increased by one reduction operation.
if (Legal->getReductionVars()->size() &&
TheLoop->getLoopDepth() > 1) {
- unsigned F = static_cast<unsigned>(MaxNestedScalarReductionUF);
- SmallUF = std::min(SmallUF, F);
- StoresUF = std::min(StoresUF, F);
- LoadsUF = std::min(LoadsUF, F);
+ unsigned F = static_cast<unsigned>(MaxNestedScalarReductionIC);
+ SmallIC = std::min(SmallIC, F);
+ StoresIC = std::min(StoresIC, F);
+ LoadsIC = std::min(LoadsIC, F);
}
- if (EnableLoadStoreRuntimeUnroll && std::max(StoresUF, LoadsUF) > SmallUF) {
- DEBUG(dbgs() << "LV: Unrolling to saturate store or load ports.\n");
- return std::max(StoresUF, LoadsUF);
+ if (EnableLoadStoreRuntimeInterleave &&
+ std::max(StoresIC, LoadsIC) > SmallIC) {
+ DEBUG(dbgs() << "LV: Interleaving to saturate store or load ports.\n");
+ return std::max(StoresIC, LoadsIC);
}
- DEBUG(dbgs() << "LV: Unrolling to reduce branch cost.\n");
- return SmallUF;
+ DEBUG(dbgs() << "LV: Interleaving to reduce branch cost.\n");
+ return SmallIC;
+ }
+
+ // Interleave if this is a large loop (small loops are already dealt with by
+ // this
+ // point) that could benefit from interleaving.
+ bool HasReductions = (Legal->getReductionVars()->size() > 0);
+ if (TTI.enableAggressiveInterleaving(HasReductions)) {
+ DEBUG(dbgs() << "LV: Interleaving to expose ILP.\n");
+ return IC;
}
- DEBUG(dbgs() << "LV: Not Unrolling.\n");
+ DEBUG(dbgs() << "LV: Not Interleaving.\n");
return 1;
}
return TTI.getAddressComputationCost(VectorTy) +
TTI.getMemoryOpCost(I->getOpcode(), VectorTy, Alignment, AS);
+ // For an interleaved access, calculate the total cost of the whole
+ // interleave group.
+ if (Legal->isAccessInterleaved(I)) {
+ auto Group = Legal->getInterleavedAccessGroup(I);
+ assert(Group && "Fail to get an interleaved access group.");
+
+ // Only calculate the cost once at the insert position.
+ if (Group->getInsertPos() != I)
+ return 0;
+
+ unsigned InterleaveFactor = Group->getFactor();
+ Type *WideVecTy =
+ VectorType::get(VectorTy->getVectorElementType(),
+ VectorTy->getVectorNumElements() * InterleaveFactor);
+
+ // Holds the indices of existing members in an interleaved load group.
+ // An interleaved store group doesn't need this as it dones't allow gaps.
+ SmallVector<unsigned, 4> Indices;
+ if (LI) {
+ for (unsigned i = 0; i < InterleaveFactor; i++)
+ if (Group->getMember(i))
+ Indices.push_back(i);
+ }
+
+ // Calculate the cost of the whole interleaved group.
+ unsigned Cost = TTI.getInterleavedMemoryOpCost(
+ I->getOpcode(), WideVecTy, Group->getFactor(), Indices,
+ Group->getAlignment(), AS);
+
+ if (Group->isReverse())
+ Cost +=
+ Group->getNumMembers() *
+ TTI.getShuffleCost(TargetTransformInfo::SK_Reverse, VectorTy, 0);
+
+ // FIXME: The interleaved load group with a huge gap could be even more
+ // expensive than scalar operations. Then we could ignore such group and
+ // use scalar operations instead.
+ return Cost;
+ }
+
// Scalarized loads/stores.
int ConsecutiveStride = Legal->isConsecutivePtr(Ptr);
bool Reverse = ConsecutiveStride < 0;
- unsigned ScalarAllocatedSize = DL->getTypeAllocSize(ValTy);
- unsigned VectorElementSize = DL->getTypeStoreSize(VectorTy)/VF;
+ const DataLayout &DL = I->getModule()->getDataLayout();
+ unsigned ScalarAllocatedSize = DL.getTypeAllocSize(ValTy);
+ unsigned VectorElementSize = DL.getTypeStoreSize(VectorTy) / VF;
if (!ConsecutiveStride || ScalarAllocatedSize != VectorElementSize) {
bool IsComplexComputation =
isLikelyComplexAddressComputation(Ptr, Legal, SE, TheLoop);
// Wide load/stores.
unsigned Cost = TTI.getAddressComputationCost(VectorTy);
- Cost += TTI.getMemoryOpCost(I->getOpcode(), VectorTy, Alignment, AS);
+ if (Legal->isMaskRequired(I))
+ Cost += TTI.getMaskedMemoryOpCost(I->getOpcode(), VectorTy, Alignment,
+ AS);
+ else
+ Cost += TTI.getMemoryOpCost(I->getOpcode(), VectorTy, Alignment, AS);
if (Reverse)
Cost += TTI.getShuffleCost(TargetTransformInfo::SK_Reverse,
return TTI.getCastInstrCost(I->getOpcode(), VectorTy, SrcVecTy);
}
case Instruction::Call: {
+ bool NeedToScalarize;
CallInst *CI = cast<CallInst>(I);
- Intrinsic::ID ID = getIntrinsicIDForCall(CI, TLI);
- assert(ID && "Not an intrinsic call!");
- Type *RetTy = ToVectorTy(CI->getType(), VF);
- SmallVector<Type*, 4> Tys;
- for (unsigned i = 0, ie = CI->getNumArgOperands(); i != ie; ++i)
- Tys.push_back(ToVectorTy(CI->getArgOperand(i)->getType(), VF));
- return TTI.getIntrinsicInstrCost(ID, RetTy, Tys);
+ unsigned CallCost = getVectorCallCost(CI, VF, TTI, TLI, NeedToScalarize);
+ if (getIntrinsicIDForCall(CI, TLI))
+ return std::min(CallCost, getVectorIntrinsicCost(CI, VF, TTI, TLI));
+ return CallCost;
}
default: {
// We are scalarizing the instruction. Return the cost of the scalar
}// end of switch.
}
-Type* LoopVectorizationCostModel::ToVectorTy(Type *Scalar, unsigned VF) {
- if (Scalar->isVoidTy() || VF == 1)
- return Scalar;
- return VectorType::get(Scalar, VF);
-}
-
char LoopVectorize::ID = 0;
static const char lv_name[] = "Loop Vectorization";
INITIALIZE_PASS_BEGIN(LoopVectorize, LV_NAME, lv_name, false, false)
INITIALIZE_PASS_DEPENDENCY(TargetTransformInfoWrapperPass)
INITIALIZE_AG_DEPENDENCY(AliasAnalysis)
INITIALIZE_PASS_DEPENDENCY(AssumptionCacheTracker)
-INITIALIZE_PASS_DEPENDENCY(BlockFrequencyInfo)
+INITIALIZE_PASS_DEPENDENCY(BlockFrequencyInfoWrapperPass)
INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
-INITIALIZE_PASS_DEPENDENCY(ScalarEvolution)
+INITIALIZE_PASS_DEPENDENCY(ScalarEvolutionWrapperPass)
INITIALIZE_PASS_DEPENDENCY(LCSSA)
INITIALIZE_PASS_DEPENDENCY(LoopInfoWrapperPass)
INITIALIZE_PASS_DEPENDENCY(LoopSimplify)
+INITIALIZE_PASS_DEPENDENCY(LoopAccessAnalysis)
INITIALIZE_PASS_END(LoopVectorize, LV_NAME, lv_name, false, false)
namespace llvm {
LoopVectorBody.push_back(NewIfBlock);
VectorLp->addBasicBlockToLoop(NewIfBlock, *LI);
Builder.SetInsertPoint(InsertPt);
- Instruction *OldBr = IfBlock->getTerminator();
- BranchInst::Create(CondBlock, NewIfBlock, Cmp, OldBr);
- OldBr->eraseFromParent();
+ ReplaceInstWithInst(IfBlock->getTerminator(),
+ BranchInst::Create(CondBlock, NewIfBlock, Cmp));
IfBlock = NewIfBlock;
}
}
projects / oota-llvm.git / blobdiff