LogicalResult matchAndRewrite(
      mhlo::ScatterOp scatter_op, OpAdaptor adaptor,
      ConversionPatternRewriter& rewriter) const final {
    {
      OperandRange operands = scatter_op.getInputs();
      Value indices = scatter_op.getScatterIndices();
      OperandRange updates = scatter_op.getUpdates();
      if (operands.size() != 1 || updates.size() != 1) return failure();

    }

    @Override
    public final boolean isBuildable() {
        return getBuildAbility().isBuildable();
    }

    @Override
    public DomainObjectSet<LanguageSourceSet> getInputs() {
        return inputSourceSets;
    }

    @Override
    public ModelMap<LanguageSourceSet> getSources() {
        return ModelMaps.toView(sources, LANGUAGE_SOURCE_SET_MODELTYPE);
    }

      rewriter.create<func::FuncOp>(input_op->getLoc(), func_name, func_type);

  OpBuilder::InsertionGuard guard = OpBuilder::InsertionGuard(rewriter);
  ArrayRef<Type> inputs = quantization_func.getFunctionType().getInputs();
  Block* block = rewriter.createBlock(
      &quantization_func.getBody(), quantization_func.begin(), inputs,
      SmallVector<Location>(inputs.size(), quantization_func.getLoc()));
  func_input_arg = block->getArgument(0);

                task.onlyIf("Test executable installation directory exists", element -> binary.getInstallDirectory().get().getAsFile().exists());
                task.getInputs()
                    .dir(binary.getInstallDirectory())
                    .withPropertyName("installDirectory");
                task.setExecutable(installTask.getRunScriptFile().get().getAsFile());

    public PublishToIvyRepository() {

        // Allow the publication to participate in incremental build
        getInputs().files((Callable<FileCollection>) () -> {
            IvyPublicationInternal publicationInternal = getPublicationInternal();

                binaries.withType(BinarySpecInternal.class, new Action<BinarySpecInternal>() {
                    @Override
                    public void execute(BinarySpecInternal binary) {
                        binary.getInputs().addAll(getSources().values());
                    }
                });
            }
        }

        @Hidden @Model

          rewriter.getStringAttr("TPU"));
    }

    constexpr int64_t kDefaultCostEstimate = 1000000;
    rewriter.replaceOpWithNewOp<TF::XlaHostComputeOp>(
        op, op.getResultTypes(), op.getInputs(),
        /*ancestors=*/rewriter.getArrayAttr({}),
        rewriter.getArrayAttr(shape_attrs),
        /*shape_inference_graph=*/
        cloned_func ? SymbolRefAttr::get(cloned_func) : SymbolRefAttr(),

    llvm::function_ref<void(ArrayRef<BlockArgument>)> handle_new_size_vars =
        nullptr) {
  auto new_input_types = llvm::to_vector<8>(func.getFunctionType().getInputs());
  auto size_var_type = GetSizeVarType(OpBuilder(func));
  int64_t original_arg_count = new_input_types.size();
  for (int64_t i = 0; i < original_arg_count; ++i) {
    auto stack_type = arg_to_stack_type(i);

    // 3. Captures inputs.
    // The number of dataset inputs can be indirectly determined to be
    // total_number_of_inputs - state_inputs - captured_inputs.
    auto func_inputs = reduce_func.getFunctionType().getInputs();
    const int func_input_size = func_inputs.size();
    const int argument_size =
        reduce_dataset->getAttrOfType<ArrayAttr>("Targuments").size();
    const int state_size =

  uint64 hash(int32_t index) const;

  bool operator==(const Operation& other) const { return node_ == other.node_; }

 private:
  typedef std::vector<std::pair<Node*, int32>> Inputs;
  static Inputs GetInputs(Node* node);

  Inputs inputs_;
  Node* node_;
};

/// Represents a tensor value produced by an Operation.
class Output {
 public:
  Output() = default;
  explicit Output(Node* n) : op_(n) {}