Subversion Repositories QNX 8.QNX8 LLVM/Clang compiler suite

//===- llvm/Analysis/VectorUtils.h - Vector utilities -----------*- C++ -*-===//

//

// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.

// See https://llvm.org/LICENSE.txt for license information.

// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception

//

//===----------------------------------------------------------------------===//

//

// This file defines some vectorizer utilities.

//

//===----------------------------------------------------------------------===//

#ifndef LLVM_ANALYSIS_VECTORUTILS_H

#define LLVM_ANALYSIS_VECTORUTILS_H

#include "llvm/ADT/MapVector.h"

#include "llvm/ADT/SmallVector.h"

#include "llvm/Analysis/LoopAccessAnalysis.h"

#include "llvm/Support/CheckedArithmetic.h"

namespace llvm {

class TargetLibraryInfo;

/// Describes the type of Parameters

enum class VFParamKind {

  Vector,            // No semantic information.

  OMP_Linear,        // declare simd linear(i)

  OMP_LinearRef,     // declare simd linear(ref(i))

  OMP_LinearVal,     // declare simd linear(val(i))

  OMP_LinearUVal,    // declare simd linear(uval(i))

  OMP_LinearPos,     // declare simd linear(i:c) uniform(c)

  OMP_LinearValPos,  // declare simd linear(val(i:c)) uniform(c)

  OMP_LinearRefPos,  // declare simd linear(ref(i:c)) uniform(c)

  OMP_LinearUValPos, // declare simd linear(uval(i:c)) uniform(c)

  OMP_Uniform,       // declare simd uniform(i)

  GlobalPredicate,   // Global logical predicate that acts on all lanes

                     // of the input and output mask concurrently. For

                     // example, it is implied by the `M` token in the

                     // Vector Function ABI mangled name.

  Unknown

};

/// Describes the type of Instruction Set Architecture

enum class VFISAKind {

  AdvancedSIMD, // AArch64 Advanced SIMD (NEON)

  SVE,          // AArch64 Scalable Vector Extension

  SSE,          // x86 SSE

  AVX,          // x86 AVX

  AVX2,         // x86 AVX2

  AVX512,       // x86 AVX512

  LLVM,         // LLVM internal ISA for functions that are not

  // attached to an existing ABI via name mangling.

  Unknown // Unknown ISA

};

/// Encapsulates information needed to describe a parameter.

///

/// The description of the parameter is not linked directly to

/// OpenMP or any other vector function description. This structure

/// is extendible to handle other paradigms that describe vector

/// functions and their parameters.

struct VFParameter {

  unsigned ParamPos;         // Parameter Position in Scalar Function.

  VFParamKind ParamKind;     // Kind of Parameter.

  int LinearStepOrPos = 0;   // Step or Position of the Parameter.

  Align Alignment = Align(); // Optional alignment in bytes, defaulted to 1.

  // Comparison operator.

  bool operator==(const VFParameter &Other) const {

    return std::tie(ParamPos, ParamKind, LinearStepOrPos, Alignment) ==

           std::tie(Other.ParamPos, Other.ParamKind, Other.LinearStepOrPos,

                    Other.Alignment);

  }

};

/// Contains the information about the kind of vectorization

/// available.

///

/// This object in independent on the paradigm used to

/// represent vector functions. in particular, it is not attached to

/// any target-specific ABI.

struct VFShape {

  ElementCount VF;                        // Vectorization factor.

  SmallVector<VFParameter, 8> Parameters; // List of parameter information.

  // Comparison operator.

  bool operator==(const VFShape &Other) const {

    return std::tie(VF, Parameters) == std::tie(Other.VF, Other.Parameters);

  }

  /// Update the parameter in position P.ParamPos to P.

  void updateParam(VFParameter P) {

    assert(P.ParamPos < Parameters.size() && "Invalid parameter position.");

    Parameters[P.ParamPos] = P;

    assert(hasValidParameterList() && "Invalid parameter list");

  }

  // Retrieve the VFShape that can be used to map a (scalar) function to itself,

  // with VF = 1.

  static VFShape getScalarShape(const CallInst &CI) {

    return VFShape::get(CI, ElementCount::getFixed(1),

                        /*HasGlobalPredicate*/ false);

  }

  // Retrieve the basic vectorization shape of the function, where all

  // parameters are mapped to VFParamKind::Vector with \p EC

  // lanes. Specifies whether the function has a Global Predicate

  // argument via \p HasGlobalPred.

  static VFShape get(const CallInst &CI, ElementCount EC, bool HasGlobalPred) {

    SmallVector<VFParameter, 8> Parameters;

    for (unsigned I = 0; I < CI.arg_size(); ++I)

      Parameters.push_back(VFParameter({I, VFParamKind::Vector}));

    if (HasGlobalPred)

      Parameters.push_back(

          VFParameter({CI.arg_size(), VFParamKind::GlobalPredicate}));

    return {EC, Parameters};

  }

  /// Validation check on the Parameters in the VFShape.

  bool hasValidParameterList() const;

};

/// Holds the VFShape for a specific scalar to vector function mapping.

struct VFInfo {

  VFShape Shape;          /// Classification of the vector function.

  std::string ScalarName; /// Scalar Function Name.

  std::string VectorName; /// Vector Function Name associated to this VFInfo.

  VFISAKind ISA;          /// Instruction Set Architecture.

};

namespace VFABI {

/// LLVM Internal VFABI ISA token for vector functions.

static constexpr char const *_LLVM_ = "_LLVM_";

/// Prefix for internal name redirection for vector function that

/// tells the compiler to scalarize the call using the scalar name

/// of the function. For example, a mangled name like

/// `_ZGV_LLVM_N2v_foo(_LLVM_Scalarize_foo)` would tell the

/// vectorizer to vectorize the scalar call `foo`, and to scalarize

/// it once vectorization is done.

static constexpr char const *_LLVM_Scalarize_ = "_LLVM_Scalarize_";

/// Function to construct a VFInfo out of a mangled names in the

/// following format:

///

/// <VFABI_name>{(<redirection>)}

///

/// where <VFABI_name> is the name of the vector function, mangled according

/// to the rules described in the Vector Function ABI of the target vector

/// extension (or <isa> from now on). The <VFABI_name> is in the following

/// format:

///

/// _ZGV<isa><mask><vlen><parameters>_<scalarname>[(<redirection>)]

///

/// This methods support demangling rules for the following <isa>:

///

/// * AArch64: https://developer.arm.com/docs/101129/latest

///

/// * x86 (libmvec): https://sourceware.org/glibc/wiki/libmvec and

///  https://sourceware.org/glibc/wiki/libmvec?action=AttachFile&do=view&target=VectorABI.txt

///

/// \param MangledName -> input string in the format

/// _ZGV<isa><mask><vlen><parameters>_<scalarname>[(<redirection>)].

/// \param M -> Module used to retrieve informations about the vector

/// function that are not possible to retrieve from the mangled

/// name. At the moment, this parameter is needed only to retrieve the

/// Vectorization Factor of scalable vector functions from their

/// respective IR declarations.

std::optional<VFInfo> tryDemangleForVFABI(StringRef MangledName,

                                          const Module &M);

/// This routine mangles the given VectorName according to the LangRef

/// specification for vector-function-abi-variant attribute and is specific to

/// the TLI mappings. It is the responsibility of the caller to make sure that

/// this is only used if all parameters in the vector function are vector type.

/// This returned string holds scalar-to-vector mapping:

///    _ZGV<isa><mask><vlen><vparams>_<scalarname>(<vectorname>)

///

/// where:

///

/// <isa> = "_LLVM_"

/// <mask> = "N". Note: TLI does not support masked interfaces.

/// <vlen> = Number of concurrent lanes, stored in the `VectorizationFactor`

///          field of the `VecDesc` struct. If the number of lanes is scalable

///          then 'x' is printed instead.

/// <vparams> = "v", as many as are the numArgs.

/// <scalarname> = the name of the scalar function.

/// <vectorname> = the name of the vector function.

std::string mangleTLIVectorName(StringRef VectorName, StringRef ScalarName,

                                unsigned numArgs, ElementCount VF);

/// Retrieve the `VFParamKind` from a string token.

VFParamKind getVFParamKindFromString(const StringRef Token);

// Name of the attribute where the variant mappings are stored.

static constexpr char const *MappingsAttrName = "vector-function-abi-variant";

/// Populates a set of strings representing the Vector Function ABI variants

/// associated to the CallInst CI. If the CI does not contain the

/// vector-function-abi-variant attribute, we return without populating

/// VariantMappings, i.e. callers of getVectorVariantNames need not check for

/// the presence of the attribute (see InjectTLIMappings).

void getVectorVariantNames(const CallInst &CI,

                           SmallVectorImpl<std::string> &VariantMappings);

} // end namespace VFABI

/// The Vector Function Database.

///

/// Helper class used to find the vector functions associated to a

/// scalar CallInst.

class VFDatabase {

  /// The Module of the CallInst CI.

  const Module *M;

  /// The CallInst instance being queried for scalar to vector mappings.

  const CallInst &CI;

  /// List of vector functions descriptors associated to the call

  /// instruction.

  const SmallVector<VFInfo, 8> ScalarToVectorMappings;

  /// Retrieve the scalar-to-vector mappings associated to the rule of

  /// a vector Function ABI.

  static void getVFABIMappings(const CallInst &CI,

                               SmallVectorImpl<VFInfo> &Mappings) {

    if (!CI.getCalledFunction())

      return;

    const StringRef ScalarName = CI.getCalledFunction()->getName();

    SmallVector<std::string, 8> ListOfStrings;

    // The check for the vector-function-abi-variant attribute is done when

    // retrieving the vector variant names here.

    VFABI::getVectorVariantNames(CI, ListOfStrings);

    if (ListOfStrings.empty())

      return;

    for (const auto &MangledName : ListOfStrings) {

      const std::optional<VFInfo> Shape =

          VFABI::tryDemangleForVFABI(MangledName, *(CI.getModule()));

      // A match is found via scalar and vector names, and also by

      // ensuring that the variant described in the attribute has a

      // corresponding definition or declaration of the vector

      // function in the Module M.

      if (Shape && (Shape->ScalarName == ScalarName)) {

        assert(CI.getModule()->getFunction(Shape->VectorName) &&

               "Vector function is missing.");

        Mappings.push_back(*Shape);

      }

    }

  }

public:

  /// Retrieve all the VFInfo instances associated to the CallInst CI.

  static SmallVector<VFInfo, 8> getMappings(const CallInst &CI) {

    SmallVector<VFInfo, 8> Ret;

    // Get mappings from the Vector Function ABI variants.

    getVFABIMappings(CI, Ret);

    // Other non-VFABI variants should be retrieved here.

    return Ret;

  }

  /// Constructor, requires a CallInst instance.

  VFDatabase(CallInst &CI)

      : M(CI.getModule()), CI(CI),

        ScalarToVectorMappings(VFDatabase::getMappings(CI)) {}

  /// \defgroup VFDatabase query interface.

  ///

  /// @{

  /// Retrieve the Function with VFShape \p Shape.

  Function *getVectorizedFunction(const VFShape &Shape) const {

    if (Shape == VFShape::getScalarShape(CI))

      return CI.getCalledFunction();

    for (const auto &Info : ScalarToVectorMappings)

      if (Info.Shape == Shape)

        return M->getFunction(Info.VectorName);

    return nullptr;

  }

  /// @}

};

template <typename T> class ArrayRef;

class DemandedBits;

class GetElementPtrInst;

template <typename InstTy> class InterleaveGroup;

class IRBuilderBase;

class Loop;

class ScalarEvolution;

class TargetTransformInfo;

class Type;

class Value;

namespace Intrinsic {

typedef unsigned ID;

}

/// A helper function for converting Scalar types to vector types. If

/// the incoming type is void, we return void. If the EC represents a

/// scalar, we return the scalar type.

inline Type *ToVectorTy(Type *Scalar, ElementCount EC) {

  if (Scalar->isVoidTy() || Scalar->isMetadataTy() || EC.isScalar())

    return Scalar;

  return VectorType::get(Scalar, EC);

}

inline Type *ToVectorTy(Type *Scalar, unsigned VF) {

  return ToVectorTy(Scalar, ElementCount::getFixed(VF));

}

/// Identify if the intrinsic is trivially vectorizable.

/// This method returns true if the intrinsic's argument types are all scalars

/// for the scalar form of the intrinsic and all vectors (or scalars handled by

/// isVectorIntrinsicWithScalarOpAtArg) for the vector form of the intrinsic.

bool isTriviallyVectorizable(Intrinsic::ID ID);

/// Identifies if the vector form of the intrinsic has a scalar operand.

bool isVectorIntrinsicWithScalarOpAtArg(Intrinsic::ID ID,

                                        unsigned ScalarOpdIdx);

/// Identifies if the vector form of the intrinsic has a operand that has

/// an overloaded type.

bool isVectorIntrinsicWithOverloadTypeAtArg(Intrinsic::ID ID, unsigned OpdIdx);

/// Returns intrinsic ID for call.

/// For the input call instruction it finds mapping intrinsic and returns

/// its intrinsic ID, in case it does not found it return not_intrinsic.

Intrinsic::ID getVectorIntrinsicIDForCall(const CallInst *CI,

                                          const TargetLibraryInfo *TLI);

/// 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.

unsigned getGEPInductionOperand(const GetElementPtrInst *Gep);

/// If the argument is a GEP, then returns the operand identified by

/// getGEPInductionOperand. However, if there is some other non-loop-invariant

/// operand, it returns that instead.

Value *stripGetElementPtr(Value *Ptr, ScalarEvolution *SE, Loop *Lp);

/// If a value has only one user that is a CastInst, return it.

Value *getUniqueCastUse(Value *Ptr, Loop *Lp, Type *Ty);

/// Get the stride of a pointer access in a loop. Looks for symbolic

/// strides "a[i*stride]". Returns the symbolic stride, or null otherwise.

Value *getStrideFromPointer(Value *Ptr, ScalarEvolution *SE, Loop *Lp);

/// Given a vector and an element number, see if the scalar value is

/// already around as a register, for example if it were inserted then extracted

/// from the vector.

Value *findScalarElement(Value *V, unsigned EltNo);

/// If all non-negative \p Mask elements are the same value, return that value.

/// If all elements are negative (undefined) or \p Mask contains different

/// non-negative values, return -1.

int getSplatIndex(ArrayRef<int> Mask);

/// Get splat value if the input is a splat vector or return nullptr.

/// The value may be extracted from a splat constants vector or from

/// a sequence of instructions that broadcast a single value into a vector.

Value *getSplatValue(const Value *V);

/// Return true if each element of the vector value \p V is poisoned or equal to

/// every other non-poisoned element. If an index element is specified, either

/// every element of the vector is poisoned or the element at that index is not

/// poisoned and equal to every other non-poisoned element.

/// This may be more powerful than the related getSplatValue() because it is

/// not limited by finding a scalar source value to a splatted vector.

bool isSplatValue(const Value *V, int Index = -1, unsigned Depth = 0);

/// Transform a shuffle mask's output demanded element mask into demanded

/// element masks for the 2 operands, returns false if the mask isn't valid.

/// Both \p DemandedLHS and \p DemandedRHS are initialised to [SrcWidth].

/// \p AllowUndefElts permits "-1" indices to be treated as undef.

bool getShuffleDemandedElts(int SrcWidth, ArrayRef<int> Mask,

                            const APInt &DemandedElts, APInt &DemandedLHS,

                            APInt &DemandedRHS, bool AllowUndefElts = false);

/// Replace each shuffle mask index with the scaled sequential indices for an

/// equivalent mask of narrowed elements. Mask elements that are less than 0

/// (sentinel values) are repeated in the output mask.

///

/// Example with Scale = 4:

///   <4 x i32> <3, 2, 0, -1> -->

///   <16 x i8> <12, 13, 14, 15, 8, 9, 10, 11, 0, 1, 2, 3, -1, -1, -1, -1>

///

/// This is the reverse process of widening shuffle mask elements, but it always

/// succeeds because the indexes can always be multiplied (scaled up) to map to

/// narrower vector elements.

void narrowShuffleMaskElts(int Scale, ArrayRef<int> Mask,

                           SmallVectorImpl<int> &ScaledMask);

/// Try to transform a shuffle mask by replacing elements with the scaled index

/// for an equivalent mask of widened elements. If all mask elements that would

/// map to a wider element of the new mask are the same negative number

/// (sentinel value), that element of the new mask is the same value. If any

/// element in a given slice is negative and some other element in that slice is

/// not the same value, return false (partial matches with sentinel values are

/// not allowed).

///

/// Example with Scale = 4:

///   <16 x i8> <12, 13, 14, 15, 8, 9, 10, 11, 0, 1, 2, 3, -1, -1, -1, -1> -->

///   <4 x i32> <3, 2, 0, -1>

///

/// This is the reverse process of narrowing shuffle mask elements if it

/// succeeds. This transform is not always possible because indexes may not

/// divide evenly (scale down) to map to wider vector elements.

bool widenShuffleMaskElts(int Scale, ArrayRef<int> Mask,

                          SmallVectorImpl<int> &ScaledMask);

/// Repetitively apply `widenShuffleMaskElts()` for as long as it succeeds,

/// to get the shuffle mask with widest possible elements.

void getShuffleMaskWithWidestElts(ArrayRef<int> Mask,

                                  SmallVectorImpl<int> &ScaledMask);

/// Splits and processes shuffle mask depending on the number of input and

/// output registers. The function does 2 main things: 1) splits the

/// source/destination vectors into real registers; 2) do the mask analysis to

/// identify which real registers are permuted. Then the function processes

/// resulting registers mask using provided action items. If no input register

/// is defined, \p NoInputAction action is used. If only 1 input register is

/// used, \p SingleInputAction is used, otherwise \p ManyInputsAction is used to

/// process > 2 input registers and masks.

/// \param Mask Original shuffle mask.

/// \param NumOfSrcRegs Number of source registers.

/// \param NumOfDestRegs Number of destination registers.

/// \param NumOfUsedRegs Number of actually used destination registers.

void processShuffleMasks(

    ArrayRef<int> Mask, unsigned NumOfSrcRegs, unsigned NumOfDestRegs,

    unsigned NumOfUsedRegs, function_ref<void()> NoInputAction,

    function_ref<void(ArrayRef<int>, unsigned, unsigned)> SingleInputAction,

    function_ref<void(ArrayRef<int>, unsigned, unsigned)> ManyInputsAction);

/// Compute a map of integer instructions to their minimum legal type

/// size.

///

/// C semantics force sub-int-sized values (e.g. i8, i16) to be promoted to int

/// type (e.g. i32) whenever arithmetic is performed on them.

///

/// For targets with native i8 or i16 operations, usually InstCombine can shrink

/// the arithmetic type down again. However InstCombine refuses to create

/// illegal types, so for targets without i8 or i16 registers, the lengthening

/// and shrinking remains.

///

/// Most SIMD ISAs (e.g. NEON) however support vectors of i8 or i16 even when

/// their scalar equivalents do not, so during vectorization it is important to

/// remove these lengthens and truncates when deciding the profitability of

/// vectorization.

///

/// This function analyzes the given range of instructions and determines the

/// minimum type size each can be converted to. It attempts to remove or

/// minimize type size changes across each def-use chain, so for example in the

/// following code:

///

///   %1 = load i8, i8*

///   %2 = add i8 %1, 2

///   %3 = load i16, i16*

///   %4 = zext i8 %2 to i32

///   %5 = zext i16 %3 to i32

///   %6 = add i32 %4, %5

///   %7 = trunc i32 %6 to i16

///

/// Instruction %6 must be done at least in i16, so computeMinimumValueSizes

/// will return: {%1: 16, %2: 16, %3: 16, %4: 16, %5: 16, %6: 16, %7: 16}.

///

/// If the optional TargetTransformInfo is provided, this function tries harder

/// to do less work by only looking at illegal types.

MapVector<Instruction*, uint64_t>

computeMinimumValueSizes(ArrayRef<BasicBlock*> Blocks,

                         DemandedBits &DB,

                         const TargetTransformInfo *TTI=nullptr);

/// Compute the union of two access-group lists.

///

/// If the list contains just one access group, it is returned directly. If the

/// list is empty, returns nullptr.

MDNode *uniteAccessGroups(MDNode *AccGroups1, MDNode *AccGroups2);

/// Compute the access-group list of access groups that @p Inst1 and @p Inst2

/// are both in. If either instruction does not access memory at all, it is

/// considered to be in every list.

///

/// If the list contains just one access group, it is returned directly. If the

/// list is empty, returns nullptr.

MDNode *intersectAccessGroups(const Instruction *Inst1,

                              const Instruction *Inst2);

/// Specifically, let Kinds = [MD_tbaa, MD_alias_scope, MD_noalias, MD_fpmath,

/// MD_nontemporal, MD_access_group].

/// For K in Kinds, we get the MDNode for K from each of the

/// elements of VL, compute their "intersection" (i.e., the most generic

/// metadata value that covers all of the individual values), and set I's

/// metadata for M equal to the intersection value.

///

/// This function always sets a (possibly null) value for each K in Kinds.

Instruction *propagateMetadata(Instruction *I, ArrayRef<Value *> VL);

/// Create a mask that filters the members of an interleave group where there

/// are gaps.

///

/// For example, the mask for \p Group with interleave-factor 3

/// and \p VF 4, that has only its first member present is:

///

///   <1,0,0,1,0,0,1,0,0,1,0,0>

///

/// Note: The result is a mask of 0's and 1's, as opposed to the other

/// create[*]Mask() utilities which create a shuffle mask (mask that

/// consists of indices).

Constant *createBitMaskForGaps(IRBuilderBase &Builder, unsigned VF,

                               const InterleaveGroup<Instruction> &Group);

/// Create a mask with replicated elements.

///

/// This function creates a shuffle mask for replicating each of the \p VF

/// elements in a vector \p ReplicationFactor times. It can be used to

/// transform a mask of \p VF elements into a mask of

/// \p VF * \p ReplicationFactor elements used by a predicated

/// interleaved-group of loads/stores whose Interleaved-factor ==

/// \p ReplicationFactor.

///

/// For example, the mask for \p ReplicationFactor=3 and \p VF=4 is:

///

///   <0,0,0,1,1,1,2,2,2,3,3,3>

llvm::SmallVector<int, 16> createReplicatedMask(unsigned ReplicationFactor,

                                                unsigned VF);

/// Create an interleave shuffle mask.

///

/// This function creates a shuffle mask for interleaving \p NumVecs vectors of

/// vectorization factor \p VF into a single wide vector. The mask is of the

/// form:

///

///   <0, VF, VF * 2, ..., VF * (NumVecs - 1), 1, VF + 1, VF * 2 + 1, ...>

///

/// For example, the mask for VF = 4 and NumVecs = 2 is:

///

///   <0, 4, 1, 5, 2, 6, 3, 7>.

llvm::SmallVector<int, 16> createInterleaveMask(unsigned VF, unsigned NumVecs);

/// Create a stride shuffle mask.

///

/// This function creates a shuffle mask whose elements begin at \p Start and

/// are incremented by \p Stride. The mask can be used to deinterleave an

/// interleaved vector into separate vectors of vectorization factor \p VF. The

/// mask is of the form:

///

///   <Start, Start + Stride, ..., Start + Stride * (VF - 1)>

///

/// For example, the mask for Start = 0, Stride = 2, and VF = 4 is:

///

///   <0, 2, 4, 6>

llvm::SmallVector<int, 16> createStrideMask(unsigned Start, unsigned Stride,

                                            unsigned VF);

/// Create a sequential shuffle mask.

///

/// This function creates shuffle mask whose elements are sequential and begin

/// at \p Start.  The mask contains \p NumInts integers and is padded with \p

/// NumUndefs undef values. The mask is of the form:

///

///   <Start, Start + 1, ... Start + NumInts - 1, undef_1, ... undef_NumUndefs>

///

/// For example, the mask for Start = 0, NumInsts = 4, and NumUndefs = 4 is:

///

///   <0, 1, 2, 3, undef, undef, undef, undef>

llvm::SmallVector<int, 16>

createSequentialMask(unsigned Start, unsigned NumInts, unsigned NumUndefs);

/// Given a shuffle mask for a binary shuffle, create the equivalent shuffle

/// mask assuming both operands are identical. This assumes that the unary

/// shuffle will use elements from operand 0 (operand 1 will be unused).

llvm::SmallVector<int, 16> createUnaryMask(ArrayRef<int> Mask,

                                           unsigned NumElts);

/// Concatenate a list of vectors.

///

/// This function generates code that concatenate the vectors in \p Vecs into a

/// single large vector. The number of vectors should be greater than one, and

/// their element types should be the same. The number of elements in the

/// vectors should also be the same; however, if the last vector has fewer

/// elements, it will be padded with undefs.

Value *concatenateVectors(IRBuilderBase &Builder, ArrayRef<Value *> Vecs);

/// Given a mask vector of i1, Return true if all of the elements of this

/// predicate mask are known to be false or undef.  That is, return true if all

/// lanes can be assumed inactive.

bool maskIsAllZeroOrUndef(Value *Mask);

/// Given a mask vector of i1, Return true if all of the elements of this

/// predicate mask are known to be true or undef.  That is, return true if all

/// lanes can be assumed active.

bool maskIsAllOneOrUndef(Value *Mask);

/// Given a mask vector of the form <Y x i1>, return an APInt (of bitwidth Y)

/// for each lane which may be active.

APInt possiblyDemandedEltsInMask(Value *Mask);

/// 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.

///

/// 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

///          ...

///        }

///

///      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

///        }

///

/// Note: the interleaved load group could have gaps (missing members), but

/// the interleaved store group doesn't allow gaps.

template <typename InstTy> class InterleaveGroup {

public:

  InterleaveGroup(uint32_t Factor, bool Reverse, Align Alignment)

      : Factor(Factor), Reverse(Reverse), Alignment(Alignment),

        InsertPos(nullptr) {}

  InterleaveGroup(InstTy *Instr, int32_t Stride, Align Alignment)

      : Alignment(Alignment), InsertPos(Instr) {

    Factor = std::abs(Stride);

    assert(Factor > 1 && "Invalid interleave factor");

    Reverse = Stride < 0;

    Members[0] = Instr;

  }

  bool isReverse() const { return Reverse; }

  uint32_t getFactor() const { return Factor; }

  Align getAlign() const { return Alignment; }

  uint32_t getNumMembers() const { return Members.size(); }

  /// 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(InstTy *Instr, int32_t Index, Align NewAlign) {

    // Make sure the key fits in an int32_t.

    std::optional<int32_t> MaybeKey = checkedAdd(Index, SmallestKey);

    if (!MaybeKey)

      return false;

    int32_t Key = *MaybeKey;

    // Skip if the key is used for either the tombstone or empty special values.

    if (DenseMapInfo<int32_t>::getTombstoneKey() == Key ||

        DenseMapInfo<int32_t>::getEmptyKey() == Key)

      return false;

    // Skip if there is already a member with the same index.

    if (Members.find(Key) != Members.end())

      return false;

    if (Key > LargestKey) {

      // The largest index is always less than the interleave factor.

      if (Index >= static_cast<int32_t>(Factor))

        return false;

      LargestKey = Key;

    } else if (Key < SmallestKey) {

      // Make sure the largest index fits in an int32_t.

      std::optional<int32_t> MaybeLargestIndex = checkedSub(LargestKey, Key);

      if (!MaybeLargestIndex)

        return false;

      // The largest index is always less than the interleave factor.

      if (*MaybeLargestIndex >= static_cast<int64_t>(Factor))

        return false;

      SmallestKey = Key;

    }

    // It's always safe to select the minimum alignment.

    Alignment = std::min(Alignment, NewAlign);

    Members[Key] = Instr;

    return true;

  }

  /// Get the member with the given index \p Index

  ///

  /// \returns nullptr if contains no such member.

  InstTy *getMember(uint32_t Index) const {

    int32_t Key = SmallestKey + Index;

    return Members.lookup(Key);

  }

  /// Get the index for the given member. Unlike the key in the member

  /// map, the index starts from 0.

  uint32_t getIndex(const InstTy *Instr) const {

    for (auto I : Members) {

      if (I.second == Instr)

        return I.first - SmallestKey;

    }

    llvm_unreachable("InterleaveGroup contains no such member");

  }

  InstTy *getInsertPos() const { return InsertPos; }

  void setInsertPos(InstTy *Inst) { InsertPos = Inst; }

  /// Add metadata (e.g. alias info) from the instructions in this group to \p

  /// NewInst.

  ///

  /// FIXME: this function currently does not add noalias metadata a'la

  /// addNewMedata.  To do that we need to compute the intersection of the

  /// noalias info from all members.

  void addMetadata(InstTy *NewInst) const;

  /// Returns true if this Group requires a scalar iteration to handle gaps.

  bool requiresScalarEpilogue() const {

    // If the last member of the Group exists, then a scalar epilog is not

    // needed for this group.

    if (getMember(getFactor() - 1))

      return false;

    // We have a group with gaps. It therefore can't be a reversed access,

    // because such groups get invalidated (TODO).

    assert(!isReverse() && "Group should have been invalidated");

    // This is a group of loads, with gaps, and without a last-member

    return true;

  }

private:

  uint32_t Factor; // Interleave Factor.

  bool Reverse;

  Align Alignment;

  DenseMap<int32_t, InstTy *> Members;

  int32_t SmallestKey = 0;

  int32_t LargestKey = 0;

  // 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

  InstTy *InsertPos;

};

/// 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(PredicatedScalarEvolution &PSE, Loop *L,

                        DominatorTree *DT, LoopInfo *LI,

                        const LoopAccessInfo *LAI)

      : PSE(PSE), TheLoop(L), DT(DT), LI(LI), LAI(LAI) {}

  ~InterleavedAccessInfo() { invalidateGroups(); }

  /// Analyze the interleaved accesses and collect them in interleave

  /// groups. Substitute symbolic strides using \p Strides.

  /// Consider also predicated loads/stores in the analysis if

  /// \p EnableMaskedInterleavedGroup is true.

  void analyzeInterleaving(bool EnableMaskedInterleavedGroup);

  /// Invalidate groups, e.g., in case all blocks in loop will be predicated

  /// contrary to original assumption. Although we currently prevent group

  /// formation for predicated accesses, we may be able to relax this limitation

  /// in the future once we handle more complicated blocks. Returns true if any

  /// groups were invalidated.

  bool invalidateGroups() {

    if (InterleaveGroups.empty()) {

      assert(

          !RequiresScalarEpilogue &&

          "RequiresScalarEpilog should not be set without interleave groups");

      return false;

    }

    InterleaveGroupMap.clear();

    for (auto *Ptr : InterleaveGroups)

      delete Ptr;

    InterleaveGroups.clear();

    RequiresScalarEpilogue = false;

    return true;

  }

  /// Check if \p Instr belongs to any interleave group.

  bool isInterleaved(Instruction *Instr) const {

    return InterleaveGroupMap.find(Instr) != InterleaveGroupMap.end();

  }

  /// Get the interleave group that \p Instr belongs to.

  ///

  /// \returns nullptr if doesn't have such group.

  InterleaveGroup<Instruction> *

  getInterleaveGroup(const Instruction *Instr) const {

    return InterleaveGroupMap.lookup(Instr);

  }

  iterator_range<SmallPtrSetIterator<llvm::InterleaveGroup<Instruction> *>>

  getInterleaveGroups() {

    return make_range(InterleaveGroups.begin(), InterleaveGroups.end());

  }

  /// Returns true if an interleaved group that may access memory

  /// out-of-bounds requires a scalar epilogue iteration for correctness.

  bool requiresScalarEpilogue() const { return RequiresScalarEpilogue; }

  /// Invalidate groups that require a scalar epilogue (due to gaps). This can

  /// happen when optimizing for size forbids a scalar epilogue, and the gap

  /// cannot be filtered by masking the load/store.

  void invalidateGroupsRequiringScalarEpilogue();

  /// Returns true if we have any interleave groups.

  bool hasGroups() const { return !InterleaveGroups.empty(); }

private:

  /// A wrapper around ScalarEvolution, used to add runtime SCEV checks.

  /// Simplifies SCEV expressions in the context of existing SCEV assumptions.

  /// The interleaved access analysis can also add new predicates (for example

  /// by versioning strides of pointers).

  PredicatedScalarEvolution &PSE;

  Loop *TheLoop;

  DominatorTree *DT;

  LoopInfo *LI;

  const LoopAccessInfo *LAI;

  /// True if the loop may contain non-reversed interleaved groups with

  /// out-of-bounds accesses. We ensure we don't speculatively access memory

  /// out-of-bounds by executing at least one scalar epilogue iteration.

  bool RequiresScalarEpilogue = false;

  /// Holds the relationships between the members and the interleave group.

  DenseMap<Instruction *, InterleaveGroup<Instruction> *> InterleaveGroupMap;

  SmallPtrSet<InterleaveGroup<Instruction> *, 4> InterleaveGroups;

  /// Holds dependences among the memory accesses in the loop. It maps a source

  /// access to a set of dependent sink accesses.

  DenseMap<Instruction *, SmallPtrSet<Instruction *, 2>> Dependences;

  /// The descriptor for a strided memory access.

  struct StrideDescriptor {

    StrideDescriptor() = default;

    StrideDescriptor(int64_t Stride, const SCEV *Scev, uint64_t Size,

                     Align Alignment)

        : Stride(Stride), Scev(Scev), Size(Size), Alignment(Alignment) {}