Codebase Browser
llvm
Go to App

llvm/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp 

//===- VPlanRecipes.cpp - Implementations for VPlan recipes ---------------===//
//
// 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
//
//===----------------------------------------------------------------------===//
///
/// \file
/// This file contains implementations for different VPlan recipes.
///
//===----------------------------------------------------------------------===//

#include "VPlan.h"
#include "VPlanAnalysis.h"
#include "VPlanUtils.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Twine.h"
#include "llvm/Analysis/IVDescriptors.h"
#include "llvm/IR/BasicBlock.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/Instruction.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/Type.h"
#include "llvm/IR/Value.h"
#include "llvm/IR/VectorBuilder.h"
#include "llvm/Support/Casting.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/LoopUtils.h"
#include "llvm/Transforms/Utils/ScalarEvolutionExpander.h"
#include <cassert>

usingnamespacellvm;

VectorParts;

namespace llvm {
extern cl::opt<bool> EnableVPlanNativePath;
}
extern cl::opt<unsigned> ForceTargetInstructionCost;

#define LV_NAME …
#define DEBUG_TYPE …

bool VPRecipeBase::mayWriteToMemory() const { … }

bool VPRecipeBase::mayReadFromMemory() const { … }

bool VPRecipeBase::mayHaveSideEffects() const { … }

void VPLiveOut::fixPhi(VPlan &Plan, VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPLiveOut::print(raw_ostream &O, VPSlotTracker &SlotTracker) const {
  O << "Live-out ";
  getPhi()->printAsOperand(O);
  O << " = ";
  getOperand(0)->printAsOperand(O, SlotTracker);
  O << "\n";
}
#endif

void VPRecipeBase::insertBefore(VPRecipeBase *InsertPos) { … }

void VPRecipeBase::insertBefore(VPBasicBlock &BB,
                                iplist<VPRecipeBase>::iterator I) { … }

void VPRecipeBase::insertAfter(VPRecipeBase *InsertPos) { … }

void VPRecipeBase::removeFromParent() { … }

iplist<VPRecipeBase>::iterator VPRecipeBase::eraseFromParent() { … }

void VPRecipeBase::moveAfter(VPRecipeBase *InsertPos) { … }

void VPRecipeBase::moveBefore(VPBasicBlock &BB,
                              iplist<VPRecipeBase>::iterator I) { … }

/// Return the underlying instruction to be used for computing \p R's cost via
/// the legacy cost model. Return nullptr if there's no suitable instruction.
static Instruction *getInstructionForCost(const VPRecipeBase *R) { … }

InstructionCost VPRecipeBase::cost(ElementCount VF, VPCostContext &Ctx) { … }

InstructionCost VPRecipeBase::computeCost(ElementCount VF,
                                          VPCostContext &Ctx) const { … }

FastMathFlags VPRecipeWithIRFlags::getFastMathFlags() const { … }

VPInstruction::VPInstruction(unsigned Opcode, CmpInst::Predicate Pred,
                             VPValue *A, VPValue *B, DebugLoc DL,
                             const Twine &Name)
    : … { … }

VPInstruction::VPInstruction(unsigned Opcode,
                             std::initializer_list<VPValue *> Operands,
                             FastMathFlags FMFs, DebugLoc DL, const Twine &Name)
    : … { … }

bool VPInstruction::doesGeneratePerAllLanes() const { … }

bool VPInstruction::canGenerateScalarForFirstLane() const { … }

Value *VPInstruction::generatePerLane(VPTransformState &State,
                                      const VPIteration &Lane) { … }

Value *VPInstruction::generatePerPart(VPTransformState &State, unsigned Part) { … }

bool VPInstruction::isVectorToScalar() const { … }

bool VPInstruction::isSingleScalar() const { … }

#if !defined(NDEBUG)
bool VPInstruction::isFPMathOp() const {
  // Inspired by FPMathOperator::classof. Notable differences are that we don't
  // support Call, PHI and Select opcodes here yet.
  return Opcode == Instruction::FAdd || Opcode == Instruction::FMul ||
         Opcode == Instruction::FNeg || Opcode == Instruction::FSub ||
         Opcode == Instruction::FDiv || Opcode == Instruction::FRem ||
         Opcode == Instruction::FCmp || Opcode == Instruction::Select;
}
#endif

void VPInstruction::execute(VPTransformState &State) { … }

bool VPInstruction::onlyFirstLaneUsed(const VPValue *Op) const { … }

bool VPInstruction::onlyFirstPartUsed(const VPValue *Op) const { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPInstruction::dump() const {
  VPSlotTracker SlotTracker(getParent()->getPlan());
  print(dbgs(), "", SlotTracker);
}

void VPInstruction::print(raw_ostream &O, const Twine &Indent,
                          VPSlotTracker &SlotTracker) const {
  O << Indent << "EMIT ";

  if (hasResult()) {
    printAsOperand(O, SlotTracker);
    O << " = ";
  }

  switch (getOpcode()) {
  case VPInstruction::Not:
    O << "not";
    break;
  case VPInstruction::SLPLoad:
    O << "combined load";
    break;
  case VPInstruction::SLPStore:
    O << "combined store";
    break;
  case VPInstruction::ActiveLaneMask:
    O << "active lane mask";
    break;
  case VPInstruction::ResumePhi:
    O << "resume-phi";
    break;
  case VPInstruction::ExplicitVectorLength:
    O << "EXPLICIT-VECTOR-LENGTH";
    break;
  case VPInstruction::FirstOrderRecurrenceSplice:
    O << "first-order splice";
    break;
  case VPInstruction::BranchOnCond:
    O << "branch-on-cond";
    break;
  case VPInstruction::CalculateTripCountMinusVF:
    O << "TC > VF ? TC - VF : 0";
    break;
  case VPInstruction::CanonicalIVIncrementForPart:
    O << "VF * Part +";
    break;
  case VPInstruction::BranchOnCount:
    O << "branch-on-count";
    break;
  case VPInstruction::ExtractFromEnd:
    O << "extract-from-end";
    break;
  case VPInstruction::ComputeReductionResult:
    O << "compute-reduction-result";
    break;
  case VPInstruction::LogicalAnd:
    O << "logical-and";
    break;
  case VPInstruction::PtrAdd:
    O << "ptradd";
    break;
  default:
    O << Instruction::getOpcodeName(getOpcode());
  }

  printFlags(O);
  printOperands(O, SlotTracker);

  if (auto DL = getDebugLoc()) {
    O << ", !dbg ";
    DL.print(O);
  }
}
#endif

void VPIRInstruction::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPIRInstruction::print(raw_ostream &O, const Twine &Indent,
                            VPSlotTracker &SlotTracker) const {
  O << Indent << "IR " << I;

  if (getNumOperands() != 0) {
    assert(getNumOperands() == 1 && "can have at most 1 operand");
    O << " (extra operand: ";
    printOperands(O, SlotTracker);
    O << ")";
  }
}
#endif

void VPWidenCallRecipe::execute(VPTransformState &State) { … }

InstructionCost VPWidenCallRecipe::computeCost(ElementCount VF,
                                               VPCostContext &Ctx) const { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPWidenCallRecipe::print(raw_ostream &O, const Twine &Indent,
                              VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN-CALL ";

  Function *CalledFn = getCalledScalarFunction();
  if (CalledFn->getReturnType()->isVoidTy())
    O << "void ";
  else {
    printAsOperand(O, SlotTracker);
    O << " = ";
  }

  O << "call @" << CalledFn->getName() << "(";
  interleaveComma(arg_operands(), O, [&O, &SlotTracker](VPValue *Op) {
    Op->printAsOperand(O, SlotTracker);
  });
  O << ")";

  if (VectorIntrinsicID)
    O << " (using vector intrinsic)";
  else {
    O << " (using library function";
    if (Variant->hasName())
      O << ": " << Variant->getName();
    O << ")";
  }
}

void VPWidenSelectRecipe::print(raw_ostream &O, const Twine &Indent,
                                VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN-SELECT ";
  printAsOperand(O, SlotTracker);
  O << " = select ";
  getOperand(0)->printAsOperand(O, SlotTracker);
  O << ", ";
  getOperand(1)->printAsOperand(O, SlotTracker);
  O << ", ";
  getOperand(2)->printAsOperand(O, SlotTracker);
  O << (isInvariantCond() ? " (condition is loop invariant)" : "");
}
#endif

void VPWidenSelectRecipe::execute(VPTransformState &State) { … }

VPRecipeWithIRFlags::FastMathFlagsTy::FastMathFlagsTy(
    const FastMathFlags &FMF) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPRecipeWithIRFlags::printFlags(raw_ostream &O) const {
  switch (OpType) {
  case OperationType::Cmp:
    O << " " << CmpInst::getPredicateName(getPredicate());
    break;
  case OperationType::DisjointOp:
    if (DisjointFlags.IsDisjoint)
      O << " disjoint";
    break;
  case OperationType::PossiblyExactOp:
    if (ExactFlags.IsExact)
      O << " exact";
    break;
  case OperationType::OverflowingBinOp:
    if (WrapFlags.HasNUW)
      O << " nuw";
    if (WrapFlags.HasNSW)
      O << " nsw";
    break;
  case OperationType::FPMathOp:
    getFastMathFlags().print(O);
    break;
  case OperationType::GEPOp:
    if (GEPFlags.IsInBounds)
      O << " inbounds";
    break;
  case OperationType::NonNegOp:
    if (NonNegFlags.NonNeg)
      O << " nneg";
    break;
  case OperationType::Other:
    break;
  }
  if (getNumOperands() > 0)
    O << " ";
}
#endif

void VPWidenRecipe::execute(VPTransformState &State) { … }

InstructionCost VPWidenRecipe::computeCost(ElementCount VF,
                                           VPCostContext &Ctx) const { … }

void VPWidenEVLRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPWidenRecipe::print(raw_ostream &O, const Twine &Indent,
                          VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN ";
  printAsOperand(O, SlotTracker);
  O << " = " << Instruction::getOpcodeName(Opcode);
  printFlags(O);
  printOperands(O, SlotTracker);
}

void VPWidenEVLRecipe::print(raw_ostream &O, const Twine &Indent,
                             VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN-VP ";
  printAsOperand(O, SlotTracker);
  O << " = " << Instruction::getOpcodeName(getOpcode());
  printFlags(O);
  printOperands(O, SlotTracker);
}
#endif

void VPWidenCastRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPWidenCastRecipe::print(raw_ostream &O, const Twine &Indent,
                              VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN-CAST ";
  printAsOperand(O, SlotTracker);
  O << " = " << Instruction::getOpcodeName(Opcode) << " ";
  printFlags(O);
  printOperands(O, SlotTracker);
  O << " to " << *getResultType();
}
#endif

/// This function adds
/// (StartIdx * Step, (StartIdx + 1) * Step, (StartIdx + 2) * Step, ...)
/// to each vector element of Val. The sequence starts at StartIndex.
/// \p Opcode is relevant for FP induction variable.
static Value *getStepVector(Value *Val, Value *StartIdx, Value *Step,
                            Instruction::BinaryOps BinOp, ElementCount VF,
                            IRBuilderBase &Builder) { … }

/// A helper function that returns an integer or floating-point constant with
/// value C.
static Constant *getSignedIntOrFpConstant(Type *Ty, int64_t C) { … }

void VPWidenIntOrFpInductionRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPWidenIntOrFpInductionRecipe::print(raw_ostream &O, const Twine &Indent,
                                          VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN-INDUCTION";
  if (getTruncInst()) {
    O << "\\l\"";
    O << " +\n" << Indent << "\"  " << VPlanIngredient(IV) << "\\l\"";
    O << " +\n" << Indent << "\"  ";
    getVPValue(0)->printAsOperand(O, SlotTracker);
  } else
    O << " " << VPlanIngredient(IV);

  O << ", ";
  getStepValue()->printAsOperand(O, SlotTracker);

  O << ", ";
  getVFValue()->printAsOperand(O, SlotTracker);
}
#endif

bool VPWidenIntOrFpInductionRecipe::isCanonical() const { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPDerivedIVRecipe::print(raw_ostream &O, const Twine &Indent,
                              VPSlotTracker &SlotTracker) const {
  O << Indent;
  printAsOperand(O, SlotTracker);
  O << " = DERIVED-IV ";
  getStartValue()->printAsOperand(O, SlotTracker);
  O << " + ";
  getOperand(1)->printAsOperand(O, SlotTracker);
  O << " * ";
  getStepValue()->printAsOperand(O, SlotTracker);
}
#endif

void VPScalarIVStepsRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPScalarIVStepsRecipe::print(raw_ostream &O, const Twine &Indent,
                                  VPSlotTracker &SlotTracker) const {
  O << Indent;
  printAsOperand(O, SlotTracker);
  O << " = SCALAR-STEPS ";
  printOperands(O, SlotTracker);
}
#endif

void VPWidenGEPRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPWidenGEPRecipe::print(raw_ostream &O, const Twine &Indent,
                             VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN-GEP ";
  O << (isPointerLoopInvariant() ? "Inv" : "Var");
  for (size_t I = 0; I < getNumOperands() - 1; ++I)
    O << "[" << (isIndexLoopInvariant(I) ? "Inv" : "Var") << "]";

  O << " ";
  printAsOperand(O, SlotTracker);
  O << " = getelementptr";
  printFlags(O);
  printOperands(O, SlotTracker);
}
#endif

void VPVectorPointerRecipe ::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPVectorPointerRecipe::print(raw_ostream &O, const Twine &Indent,
                                  VPSlotTracker &SlotTracker) const {
  O << Indent;
  printAsOperand(O, SlotTracker);
  O << " = vector-pointer ";
  if (IsReverse)
    O << "(reverse) ";

  printOperands(O, SlotTracker);
}
#endif

void VPBlendRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPBlendRecipe::print(raw_ostream &O, const Twine &Indent,
                          VPSlotTracker &SlotTracker) const {
  O << Indent << "BLEND ";
  printAsOperand(O, SlotTracker);
  O << " =";
  if (getNumIncomingValues() == 1) {
    // Not a User of any mask: not really blending, this is a
    // single-predecessor phi.
    O << " ";
    getIncomingValue(0)->printAsOperand(O, SlotTracker);
  } else {
    for (unsigned I = 0, E = getNumIncomingValues(); I < E; ++I) {
      O << " ";
      getIncomingValue(I)->printAsOperand(O, SlotTracker);
      if (I == 0)
        continue;
      O << "/";
      getMask(I)->printAsOperand(O, SlotTracker);
    }
  }
}
#endif

void VPReductionRecipe::execute(VPTransformState &State) { … }

void VPReductionEVLRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPReductionRecipe::print(raw_ostream &O, const Twine &Indent,
                              VPSlotTracker &SlotTracker) const {
  O << Indent << "REDUCE ";
  printAsOperand(O, SlotTracker);
  O << " = ";
  getChainOp()->printAsOperand(O, SlotTracker);
  O << " +";
  if (isa<FPMathOperator>(getUnderlyingInstr()))
    O << getUnderlyingInstr()->getFastMathFlags();
  O << " reduce." << Instruction::getOpcodeName(RdxDesc.getOpcode()) << " (";
  getVecOp()->printAsOperand(O, SlotTracker);
  if (isConditional()) {
    O << ", ";
    getCondOp()->printAsOperand(O, SlotTracker);
  }
  O << ")";
  if (RdxDesc.IntermediateStore)
    O << " (with final reduction value stored in invariant address sank "
         "outside of loop)";
}

void VPReductionEVLRecipe::print(raw_ostream &O, const Twine &Indent,
                                 VPSlotTracker &SlotTracker) const {
  const RecurrenceDescriptor &RdxDesc = getRecurrenceDescriptor();
  O << Indent << "REDUCE ";
  printAsOperand(O, SlotTracker);
  O << " = ";
  getChainOp()->printAsOperand(O, SlotTracker);
  O << " +";
  if (isa<FPMathOperator>(getUnderlyingInstr()))
    O << getUnderlyingInstr()->getFastMathFlags();
  O << " vp.reduce." << Instruction::getOpcodeName(RdxDesc.getOpcode()) << " (";
  getVecOp()->printAsOperand(O, SlotTracker);
  O << ", ";
  getEVL()->printAsOperand(O, SlotTracker);
  if (isConditional()) {
    O << ", ";
    getCondOp()->printAsOperand(O, SlotTracker);
  }
  O << ")";
  if (RdxDesc.IntermediateStore)
    O << " (with final reduction value stored in invariant address sank "
         "outside of loop)";
}
#endif

bool VPReplicateRecipe::shouldPack() const { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPReplicateRecipe::print(raw_ostream &O, const Twine &Indent,
                              VPSlotTracker &SlotTracker) const {
  O << Indent << (IsUniform ? "CLONE " : "REPLICATE ");

  if (!getUnderlyingInstr()->getType()->isVoidTy()) {
    printAsOperand(O, SlotTracker);
    O << " = ";
  }
  if (auto *CB = dyn_cast<CallBase>(getUnderlyingInstr())) {
    O << "call";
    printFlags(O);
    O << "@" << CB->getCalledFunction()->getName() << "(";
    interleaveComma(make_range(op_begin(), op_begin() + (getNumOperands() - 1)),
                    O, [&O, &SlotTracker](VPValue *Op) {
                      Op->printAsOperand(O, SlotTracker);
                    });
    O << ")";
  } else {
    O << Instruction::getOpcodeName(getUnderlyingInstr()->getOpcode());
    printFlags(O);
    printOperands(O, SlotTracker);
  }

  if (shouldPack())
    O << " (S->V)";
}
#endif

/// Checks if \p C is uniform across all VFs and UFs. It is considered as such
/// if it is either defined outside the vector region or its operand is known to
/// be uniform across all VFs and UFs (e.g. VPDerivedIV or VPCanonicalIVPHI).
/// TODO: Uniformity should be associated with a VPValue and there should be a
/// generic way to check.
static bool isUniformAcrossVFsAndUFs(VPScalarCastRecipe *C) { … }

Value *VPScalarCastRecipe ::generate(VPTransformState &State, unsigned Part) { … }

void VPScalarCastRecipe ::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPScalarCastRecipe ::print(raw_ostream &O, const Twine &Indent,
                                VPSlotTracker &SlotTracker) const {
  O << Indent << "SCALAR-CAST ";
  printAsOperand(O, SlotTracker);
  O << " = " << Instruction::getOpcodeName(Opcode) << " ";
  printOperands(O, SlotTracker);
  O << " to " << *ResultTy;
}
#endif

void VPBranchOnMaskRecipe::execute(VPTransformState &State) { … }

void VPPredInstPHIRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPPredInstPHIRecipe::print(raw_ostream &O, const Twine &Indent,
                                VPSlotTracker &SlotTracker) const {
  O << Indent << "PHI-PREDICATED-INSTRUCTION ";
  printAsOperand(O, SlotTracker);
  O << " = ";
  printOperands(O, SlotTracker);
}
#endif

InstructionCost VPWidenMemoryRecipe::computeCost(ElementCount VF,
                                                 VPCostContext &Ctx) const { … }

void VPWidenLoadRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPWidenLoadRecipe::print(raw_ostream &O, const Twine &Indent,
                              VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN ";
  printAsOperand(O, SlotTracker);
  O << " = load ";
  printOperands(O, SlotTracker);
}
#endif

/// Use all-true mask for reverse rather than actual mask, as it avoids a
/// dependence w/o affecting the result.
static Instruction *createReverseEVL(IRBuilderBase &Builder, Value *Operand,
                                     Value *EVL, const Twine &Name) { … }

void VPWidenLoadEVLRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPWidenLoadEVLRecipe::print(raw_ostream &O, const Twine &Indent,
                                 VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN ";
  printAsOperand(O, SlotTracker);
  O << " = vp.load ";
  printOperands(O, SlotTracker);
}
#endif

void VPWidenStoreRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPWidenStoreRecipe::print(raw_ostream &O, const Twine &Indent,
                               VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN store ";
  printOperands(O, SlotTracker);
}
#endif

void VPWidenStoreEVLRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPWidenStoreEVLRecipe::print(raw_ostream &O, const Twine &Indent,
                                  VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN vp.store ";
  printOperands(O, SlotTracker);
}
#endif

static Value *createBitOrPointerCast(IRBuilderBase &Builder, Value *V,
                                     VectorType *DstVTy, const DataLayout &DL) { … }

/// Return a vector containing interleaved elements from multiple
/// smaller input vectors.
static Value *interleaveVectors(IRBuilderBase &Builder, ArrayRef<Value *> Vals,
                                const Twine &Name) { … }

// 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, poison, <0, 3, 6, 9>   ; R elements
//   %G.vec = shuffle %wide.vec, poison, <1, 4, 7, 10>  ; G elements
//   %B.vec = shuffle %wide.vec, poison, <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, poison, <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 VPInterleaveRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPInterleaveRecipe::print(raw_ostream &O, const Twine &Indent,
                               VPSlotTracker &SlotTracker) const {
  O << Indent << "INTERLEAVE-GROUP with factor " << IG->getFactor() << " at ";
  IG->getInsertPos()->printAsOperand(O, false);
  O << ", ";
  getAddr()->printAsOperand(O, SlotTracker);
  VPValue *Mask = getMask();
  if (Mask) {
    O << ", ";
    Mask->printAsOperand(O, SlotTracker);
  }

  unsigned OpIdx = 0;
  for (unsigned i = 0; i < IG->getFactor(); ++i) {
    if (!IG->getMember(i))
      continue;
    if (getNumStoreOperands() > 0) {
      O << "\n" << Indent << "  store ";
      getOperand(1 + OpIdx)->printAsOperand(O, SlotTracker);
      O << " to index " << i;
    } else {
      O << "\n" << Indent << "  ";
      getVPValue(OpIdx)->printAsOperand(O, SlotTracker);
      O << " = load from index " << i;
    }
    ++OpIdx;
  }
}
#endif

void VPCanonicalIVPHIRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPCanonicalIVPHIRecipe::print(raw_ostream &O, const Twine &Indent,
                                   VPSlotTracker &SlotTracker) const {
  O << Indent << "EMIT ";
  printAsOperand(O, SlotTracker);
  O << " = CANONICAL-INDUCTION ";
  printOperands(O, SlotTracker);
}
#endif

bool VPCanonicalIVPHIRecipe::isCanonical(
    InductionDescriptor::InductionKind Kind, VPValue *Start,
    VPValue *Step) const { … }

bool VPWidenPointerInductionRecipe::onlyScalarsGenerated(bool IsScalable) { … }

void VPWidenPointerInductionRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPWidenPointerInductionRecipe::print(raw_ostream &O, const Twine &Indent,
                                          VPSlotTracker &SlotTracker) const {
  O << Indent << "EMIT ";
  printAsOperand(O, SlotTracker);
  O << " = WIDEN-POINTER-INDUCTION ";
  getStartValue()->printAsOperand(O, SlotTracker);
  O << ", " << *IndDesc.getStep();
}
#endif

void VPExpandSCEVRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPExpandSCEVRecipe::print(raw_ostream &O, const Twine &Indent,
                               VPSlotTracker &SlotTracker) const {
  O << Indent << "EMIT ";
  getVPSingleValue()->printAsOperand(O, SlotTracker);
  O << " = EXPAND SCEV " << *Expr;
}
#endif

void VPWidenCanonicalIVRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPWidenCanonicalIVRecipe::print(raw_ostream &O, const Twine &Indent,
                                     VPSlotTracker &SlotTracker) const {
  O << Indent << "EMIT ";
  printAsOperand(O, SlotTracker);
  O << " = WIDEN-CANONICAL-INDUCTION ";
  printOperands(O, SlotTracker);
}
#endif

void VPFirstOrderRecurrencePHIRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPFirstOrderRecurrencePHIRecipe::print(raw_ostream &O, const Twine &Indent,
                                            VPSlotTracker &SlotTracker) const {
  O << Indent << "FIRST-ORDER-RECURRENCE-PHI ";
  printAsOperand(O, SlotTracker);
  O << " = phi ";
  printOperands(O, SlotTracker);
}
#endif

void VPReductionPHIRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPReductionPHIRecipe::print(raw_ostream &O, const Twine &Indent,
                                 VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN-REDUCTION-PHI ";

  printAsOperand(O, SlotTracker);
  O << " = phi ";
  printOperands(O, SlotTracker);
}
#endif

void VPWidenPHIRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPWidenPHIRecipe::print(raw_ostream &O, const Twine &Indent,
                             VPSlotTracker &SlotTracker) const {
  O << Indent << "WIDEN-PHI ";

  auto *OriginalPhi = cast<PHINode>(getUnderlyingValue());
  // Unless all incoming values are modeled in VPlan  print the original PHI
  // directly.
  // TODO: Remove once all VPWidenPHIRecipe instances keep all relevant incoming
  // values as VPValues.
  if (getNumOperands() != OriginalPhi->getNumOperands()) {
    O << VPlanIngredient(OriginalPhi);
    return;
  }

  printAsOperand(O, SlotTracker);
  O << " = phi ";
  printOperands(O, SlotTracker);
}
#endif

// TODO: It would be good to use the existing VPWidenPHIRecipe instead and
// remove VPActiveLaneMaskPHIRecipe.
void VPActiveLaneMaskPHIRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPActiveLaneMaskPHIRecipe::print(raw_ostream &O, const Twine &Indent,
                                      VPSlotTracker &SlotTracker) const {
  O << Indent << "ACTIVE-LANE-MASK-PHI ";

  printAsOperand(O, SlotTracker);
  O << " = phi ";
  printOperands(O, SlotTracker);
}
#endif

void VPEVLBasedIVPHIRecipe::execute(VPTransformState &State) { … }

#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
void VPEVLBasedIVPHIRecipe::print(raw_ostream &O, const Twine &Indent,
                                  VPSlotTracker &SlotTracker) const {
  O << Indent << "EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI ";

  printAsOperand(O, SlotTracker);
  O << " = phi ";
  printOperands(O, SlotTracker);
}
#endif