C++ LoadInst::getAlignment方法代码示例

bool AMDGPUCodeGenPrepare::canWidenScalarExtLoad(LoadInst &I) const {
  Type *Ty = I.getType();
  const DataLayout &DL = Mod->getDataLayout();
  int TySize = DL.getTypeSizeInBits(Ty);
  unsigned Align = I.getAlignment() ?
                   I.getAlignment() : DL.getABITypeAlignment(Ty);

  return I.isSimple() && TySize < 32 && Align >= 4 && DA->isUniform(&I);
}

///   %res = load {atomic|volatile} T* %ptr memory_order, align sizeof(T)
/// becomes:
///   %res = call T @llvm.nacl.atomic.load.i<size>(%ptr, memory_order)
void AtomicVisitor::visitLoadInst(LoadInst &I) {
  return; // XXX EMSCRIPTEN
  if (I.isSimple())
    return;
  PointerHelper<LoadInst> PH(*this, I);
  const NaCl::AtomicIntrinsics::AtomicIntrinsic *Intrinsic =
      findAtomicIntrinsic(I, Intrinsic::nacl_atomic_load, PH.PET);
  checkAlignment(I, I.getAlignment(), PH.BitSize / CHAR_BIT);
  Value *Args[] = {PH.P, freezeMemoryOrder(I, I.getOrdering())};
  replaceInstructionWithIntrinsicCall(I, Intrinsic, PH.OriginalPET, PH.PET,
                                      Args);
}

void X86InterleavedAccessGroup::decompose(
    Instruction *VecInst, unsigned NumSubVectors, VectorType *SubVecTy,
    SmallVectorImpl<Instruction *> &DecomposedVectors) {

  assert((isa<LoadInst>(VecInst) || isa<ShuffleVectorInst>(VecInst)) &&
         "Expected Load or Shuffle");

  Type *VecTy = VecInst->getType();
  (void)VecTy;
  assert(VecTy->isVectorTy() &&
         DL.getTypeSizeInBits(VecTy) >=
             DL.getTypeSizeInBits(SubVecTy) * NumSubVectors &&
         "Invalid Inst-size!!!");

  if (auto *SVI = dyn_cast<ShuffleVectorInst>(VecInst)) {
    Value *Op0 = SVI->getOperand(0);
    Value *Op1 = SVI->getOperand(1);

    // Generate N(= NumSubVectors) shuffles of T(= SubVecTy) type.
    for (unsigned i = 0; i < NumSubVectors; ++i)
      DecomposedVectors.push_back(
          cast<ShuffleVectorInst>(Builder.CreateShuffleVector(
              Op0, Op1,
              createSequentialMask(Builder, Indices[i],
                                   SubVecTy->getVectorNumElements(), 0))));
    return;
  }

  // Decompose the load instruction.
  LoadInst *LI = cast<LoadInst>(VecInst);
  Type *VecBasePtrTy = SubVecTy->getPointerTo(LI->getPointerAddressSpace());
  Value *VecBasePtr;
  unsigned int NumLoads = NumSubVectors;
  // In the case of stride 3 with a vector of 32 elements load the information
  // in the following way:
  // [0,1...,VF/2-1,VF/2+VF,VF/2+VF+1,...,2VF-1]
  if (DL.getTypeSizeInBits(VecTy) == 768) {
    Type *VecTran =
        VectorType::get(Type::getInt8Ty(LI->getContext()), 16)->getPointerTo();
    VecBasePtr = Builder.CreateBitCast(LI->getPointerOperand(), VecTran);
    NumLoads = NumSubVectors * 2;
  } else
    VecBasePtr = Builder.CreateBitCast(LI->getPointerOperand(), VecBasePtrTy);
  // Generate N loads of T type.
  for (unsigned i = 0; i < NumLoads; i++) {
    // TODO: Support inbounds GEP.
    Value *NewBasePtr = Builder.CreateGEP(VecBasePtr, Builder.getInt32(i));
    Instruction *NewLoad =
        Builder.CreateAlignedLoad(NewBasePtr, LI->getAlignment());
    DecomposedVectors.push_back(NewLoad);
  }
}

/// \brief Try to aggregate loads from a sorted list of loads to be combined.
///
/// It is guaranteed that no writes occur between any of the loads. All loads
/// have the same base pointer. There are at least two loads.
bool LoadCombine::aggregateLoads(SmallVectorImpl<LoadPOPPair> &Loads) {
  assert(Loads.size() >= 2 && "Insufficient loads!");
  LoadInst *BaseLoad = nullptr;
  SmallVector<LoadPOPPair, 8> AggregateLoads;
  bool Combined = false;
  bool ValidPrevOffset = false;
  APInt PrevOffset;
  uint64_t PrevSize = 0;
  for (auto &L : Loads) {
    if (ValidPrevOffset == false) {
      BaseLoad = L.Load;
      PrevOffset = L.POP.Offset;
      PrevSize = L.Load->getModule()->getDataLayout().getTypeStoreSize(
          L.Load->getType());
      AggregateLoads.push_back(L);
      ValidPrevOffset = true;
      continue;
    }
    if (L.Load->getAlignment() > BaseLoad->getAlignment())
      continue;
    APInt PrevEnd = PrevOffset + PrevSize;
    if (L.POP.Offset.sgt(PrevEnd)) {
      // No other load will be combinable
      if (combineLoads(AggregateLoads))
        Combined = true;
      AggregateLoads.clear();
      ValidPrevOffset = false;
      continue;
    }
    if (L.POP.Offset != PrevEnd)
      // This load is offset less than the size of the last load.
      // FIXME: We may want to handle this case.
      continue;
    PrevOffset = L.POP.Offset;
    PrevSize = L.Load->getModule()->getDataLayout().getTypeStoreSize(
        L.Load->getType());
    AggregateLoads.push_back(L);
  }
  if (combineLoads(AggregateLoads))
    Combined = true;
  return Combined;
}

bool Scalarizer::visitLoadInst(LoadInst &LI) {
  if (!ScalarizeLoadStore)
    return false;
  if (!LI.isSimple())
    return false;

  VectorLayout Layout;
  if (!getVectorLayout(LI.getType(), LI.getAlignment(), Layout))
    return false;

  unsigned NumElems = Layout.VecTy->getNumElements();
  IRBuilder<> Builder(LI.getParent(), &LI);
  Scatterer Ptr = scatter(&LI, LI.getPointerOperand());
  ValueVector Res;
  Res.resize(NumElems);

  for (unsigned I = 0; I < NumElems; ++I)
    Res[I] = Builder.CreateAlignedLoad(Ptr[I], Layout.getElemAlign(I),
                                       LI.getName() + ".i" + Twine(I));
  gather(&LI, Res);
  return true;
}

void Lint::visitLoadInst(LoadInst &I) {
  visitMemoryReference(I, I.getPointerOperand(),
                       DL->getTypeStoreSize(I.getType()), I.getAlignment(),
                       I.getType(), MemRef::Read);
}

unsigned CostModelAnalysis::getInstructionCost(Instruction *I) const {
  if (!VTTI)
    return -1;

  switch (I->getOpcode()) {
  case Instruction::Ret:
  case Instruction::PHI:
  case Instruction::Br: {
    return VTTI->getCFInstrCost(I->getOpcode());
  }
  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: {
    return VTTI->getArithmeticInstrCost(I->getOpcode(), I->getType());
  }
  case Instruction::Select: {
    SelectInst *SI = cast<SelectInst>(I);
    Type *CondTy = SI->getCondition()->getType();
    return VTTI->getCmpSelInstrCost(I->getOpcode(), I->getType(), CondTy);
  }
  case Instruction::ICmp:
  case Instruction::FCmp: {
    Type *ValTy = I->getOperand(0)->getType();
    return VTTI->getCmpSelInstrCost(I->getOpcode(), ValTy);
  }
  case Instruction::Store: {
    StoreInst *SI = cast<StoreInst>(I);
    Type *ValTy = SI->getValueOperand()->getType();
    return VTTI->getMemoryOpCost(I->getOpcode(), ValTy,
                                 SI->getAlignment(),
                                 SI->getPointerAddressSpace());
  }
  case Instruction::Load: {
    LoadInst *LI = cast<LoadInst>(I);
    return VTTI->getMemoryOpCost(I->getOpcode(), I->getType(),
                                 LI->getAlignment(),
                                 LI->getPointerAddressSpace());
  }
  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: {
    Type *SrcTy = I->getOperand(0)->getType();
    return VTTI->getCastInstrCost(I->getOpcode(), I->getType(), SrcTy);
  }
  case Instruction::ExtractElement: {
    ExtractElementInst * EEI = cast<ExtractElementInst>(I);
    ConstantInt *CI = dyn_cast<ConstantInt>(I->getOperand(1));
    unsigned Idx = -1;
    if (CI)
      Idx = CI->getZExtValue();
    return VTTI->getVectorInstrCost(I->getOpcode(),
                                    EEI->getOperand(0)->getType(), Idx);
  }
  case Instruction::InsertElement: {
      InsertElementInst * IE = cast<InsertElementInst>(I);
      ConstantInt *CI = dyn_cast<ConstantInt>(IE->getOperand(2));
      unsigned Idx = -1;
      if (CI)
        Idx = CI->getZExtValue();
      return VTTI->getVectorInstrCost(I->getOpcode(),
                                      IE->getType(), Idx);
    }
  default:
    // We don't have any information on this instruction.
    return -1;
  }
}

//.........这里部分代码省略.........
    // create temporary alloca space to communicate to/from.
    alloc = makeAlloca(int8Ty, "agg.tmp", insertBefore,
                       Range.End-Range.Start, Alignment);

    // Generate the old and new base pointers before we output
    // anything else.
    {
      Type* iPtrTy = TD->getIntPtrType(alloc->getType());
      Type* iNewBaseTy = TD->getIntPtrType(alloc->getType());
      oldBaseI = builder.CreatePtrToInt(StartPtr, iPtrTy, "agg.tmp.oldb.i");
      newBaseI = builder.CreatePtrToInt(alloc, iNewBaseTy, "agg.tmp.newb.i");
    }

    // If storing, do the stores we had into our alloca'd region.
    if( isStore ) {
      for (SmallVector<Instruction*, 16>::const_iterator
           SI = Range.TheStores.begin(),
           SE = Range.TheStores.end(); SI != SE; ++SI) {
        StoreInst* oldStore = cast<StoreInst>(*SI);

        if( DebugThis ) {
          errs() << "have store in range:";
          oldStore->dump();
        }

        Value* ptrToAlloc = rebasePointer(oldStore->getPointerOperand(),
                                          StartPtr, alloc, "agg.tmp",
                                          &builder, *TD, oldBaseI, newBaseI);
        // Old load must not be volatile or atomic... or we shouldn't have put
        // it in ranges
        assert(!(oldStore->isVolatile() || oldStore->isAtomic()));
        StoreInst* newStore =
          builder.CreateStore(oldStore->getValueOperand(), ptrToAlloc);
        newStore->setAlignment(oldStore->getAlignment());
        newStore->takeName(oldStore);
      }
    }

    // cast the pointer that was load/stored to i8 if necessary.
    if( StartPtr->getType()->getPointerElementType() == int8Ty ) {
      globalPtr = StartPtr;
    } else {
      globalPtr = builder.CreatePointerCast(StartPtr, globalInt8PtrTy, "agg.cast");
    }

    // Get a Constant* for the length.
    Constant* len = ConstantInt::get(sizeTy, Range.End-Range.Start, false);

    // Now add the memcpy instruction
    unsigned addrSpaceDst,addrSpaceSrc;
    addrSpaceDst = addrSpaceSrc = 0;
    if( isStore ) addrSpaceDst = globalSpace;
    if( isLoad ) addrSpaceSrc = globalSpace;

    Type *types[3];
    types[0] = PointerType::get(int8Ty, addrSpaceDst);
    types[1] = PointerType::get(int8Ty, addrSpaceSrc);
    types[2] = sizeTy;

    Function *func = Intrinsic::getDeclaration(M, Intrinsic::memcpy, types);

    Value* args[5]; // dst src len alignment isvolatile
    if( isStore ) {
      // it's a store (ie put)
      args[0] = globalPtr;
      args[1] = alloc;

Value *BoUpSLP::vectorizeTree(ValueList &VL, int VF) {
  Type *ScalarTy = VL[0]->getType();
  if (StoreInst *SI = dyn_cast<StoreInst>(VL[0]))
    ScalarTy = SI->getValueOperand()->getType();
  VectorType *VecTy = VectorType::get(ScalarTy, VF);

  // Check if all of the operands are constants or identical.
  bool AllConst = true;
  bool AllSameScalar = true;
  for (unsigned i = 0, e = VF; i < e; ++i) {
    AllConst &= !!dyn_cast<Constant>(VL[i]);
    AllSameScalar &= (VL[0] == VL[i]);
    // Must have a single use.
    Instruction *I = dyn_cast<Instruction>(VL[i]);
    if (I && (I->getNumUses() > 1 || I->getParent() != BB))
      return Scalarize(VL, VecTy);
  }

  // Is this a simple vector constant.
  if (AllConst || AllSameScalar) return Scalarize(VL, VecTy);

  // Scalarize unknown structures.
  Instruction *VL0 = dyn_cast<Instruction>(VL[0]);
  if (!VL0) return Scalarize(VL, VecTy);

  unsigned Opcode = VL0->getOpcode();
  for (unsigned i = 0, e = VF; i < e; ++i) {
    Instruction *I = dyn_cast<Instruction>(VL[i]);
    // If not all of the instructions are identical then we have to scalarize.
    if (!I || Opcode != I->getOpcode()) return Scalarize(VL, VecTy);
  }

  switch (Opcode) {
  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: {
    ValueList LHSVL, RHSVL;
    for (int i = 0; i < VF; ++i) {
      RHSVL.push_back(cast<Instruction>(VL[i])->getOperand(0));
      LHSVL.push_back(cast<Instruction>(VL[i])->getOperand(1));
    }

    Value *RHS = vectorizeTree(RHSVL, VF);
    Value *LHS = vectorizeTree(LHSVL, VF);
    IRBuilder<> Builder(GetLastInstr(VL, VF));
    BinaryOperator *BinOp = dyn_cast<BinaryOperator>(VL0);
    return Builder.CreateBinOp(BinOp->getOpcode(), RHS,LHS);
  }
  case Instruction::Load: {
    LoadInst *LI = dyn_cast<LoadInst>(VL0);
    unsigned Alignment = LI->getAlignment();

    // Check if all of the loads are consecutive.
    for (unsigned i = 1, e = VF; i < e; ++i)
      if (!isConsecutiveAccess(VL[i-1], VL[i]))
        return Scalarize(VL, VecTy);

    IRBuilder<> Builder(GetLastInstr(VL, VF));
    Value *VecPtr = Builder.CreateBitCast(LI->getPointerOperand(),
                                          VecTy->getPointerTo());
    LI = Builder.CreateLoad(VecPtr);
    LI->setAlignment(Alignment);
    return LI;
  }
  case Instruction::Store: {
    StoreInst *SI = dyn_cast<StoreInst>(VL0);
    unsigned Alignment = SI->getAlignment();

    ValueList ValueOp;
    for (int i = 0; i < VF; ++i)
      ValueOp.push_back(cast<StoreInst>(VL[i])->getValueOperand());

    Value *VecValue = vectorizeTree(ValueOp, VF);

    IRBuilder<> Builder(GetLastInstr(VL, VF));
    Value *VecPtr = Builder.CreateBitCast(SI->getPointerOperand(),
                                          VecTy->getPointerTo());
    Builder.CreateStore(VecValue, VecPtr)->setAlignment(Alignment);

    for (int i = 0; i < VF; ++i)
      cast<Instruction>(VL[i])->eraseFromParent();
    return 0;
  }
  default:
    return Scalarize(VL, VecTy);
//.........这里部分代码省略.........

void Lint::visitLoadInst(LoadInst &I) {
  visitMemoryReference(I, I.getPointerOperand(), I.getAlignment(), I.getType());
}

//.........这里部分代码省略.........
        // Non-dead argument: insert GEPs and loads as appropriate.
        ScalarizeTable &ArgIndices = ScalarizedElements[&*I];
        // Store the Value* version of the indices in here, but declare it now
        // for reuse.
        std::vector<Value *> Ops;
        for (const auto &ArgIndex : ArgIndices) {
          Value *V = *AI;
          LoadInst *OrigLoad =
              OriginalLoads[std::make_pair(&*I, ArgIndex.second)];
          if (!ArgIndex.second.empty()) {
            Ops.reserve(ArgIndex.second.size());
            Type *ElTy = V->getType();
            for (auto II : ArgIndex.second) {
              // Use i32 to index structs, and i64 for others (pointers/arrays).
              // This satisfies GEP constraints.
              Type *IdxTy =
                  (ElTy->isStructTy() ? Type::getInt32Ty(F->getContext())
                                      : Type::getInt64Ty(F->getContext()));
              Ops.push_back(ConstantInt::get(IdxTy, II));
              // Keep track of the type we're currently indexing.
              if (auto *ElPTy = dyn_cast<PointerType>(ElTy))
                ElTy = ElPTy->getElementType();
              else
                ElTy = cast<CompositeType>(ElTy)->getTypeAtIndex(II);
            }
            // And create a GEP to extract those indices.
            V = GetElementPtrInst::Create(ArgIndex.first, V, Ops,
                                          V->getName() + ".idx", Call);
            Ops.clear();
          }
          // Since we're replacing a load make sure we take the alignment
          // of the previous load.
          LoadInst *newLoad = new LoadInst(V, V->getName() + ".val", Call);
          newLoad->setAlignment(OrigLoad->getAlignment());
          // Transfer the AA info too.
          AAMDNodes AAInfo;
          OrigLoad->getAAMetadata(AAInfo);
          newLoad->setAAMetadata(AAInfo);

          Args.push_back(newLoad);
          ArgAttrVec.push_back(AttributeSet());
        }
      }

    // Push any varargs arguments on the list.
    for (; AI != CS.arg_end(); ++AI, ++ArgNo) {
      Args.push_back(*AI);
      ArgAttrVec.push_back(CallPAL.getParamAttributes(ArgNo));
    }

    SmallVector<OperandBundleDef, 1> OpBundles;
    CS.getOperandBundlesAsDefs(OpBundles);

    CallSite NewCS;
    if (InvokeInst *II = dyn_cast<InvokeInst>(Call)) {
      NewCS = InvokeInst::Create(NF, II->getNormalDest(), II->getUnwindDest(),
                                 Args, OpBundles, "", Call);
    } else {
      auto *NewCall = CallInst::Create(NF, Args, OpBundles, "", Call);
      NewCall->setTailCallKind(cast<CallInst>(Call)->getTailCallKind());
      NewCS = NewCall;
    }
    NewCS.setCallingConv(CS.getCallingConv());
    NewCS.setAttributes(
        AttributeList::get(F->getContext(), CallPAL.getFnAttributes(),
                           CallPAL.getRetAttributes(), ArgAttrVec));