// If this is a terminator, identify the values to use to replace the
      // original If op.
      if (op_to_inline.hasTrait<OpTrait::IsTerminator>()) {
        updated_results.reserve(op_to_inline.getNumOperands());
        for (Value operand : op_to_inline.getOperands())
          updated_results.push_back(mapper.lookup(operand));
        break;
      }

      // Otherwise, clone the op here.

    if (!is_reduce_dim[dim]) {
      is_reduce_dim[dim] = true;
      num_reduce_dim++;
    }
  }

  ArrayRef<int64_t> shape = ranked_ty.getShape();
  SmallVector<int64_t, 4> out_shape;
  out_shape.reserve(rank - (keep_dims.getValue() ? 0 : num_reduce_dim));
  for (int64_t i = 0; i < rank; ++i) {
    if (!is_reduce_dim[i])
      out_shape.push_back(shape[i]);
    else if (keep_dims.getValue())
      out_shape.push_back(1);

  MlirOptimizationPassState overall_state = MlirOptimizationPassState::Disabled;

  // Cache per pass state and reuse it during pass execution.
  std::vector<MlirOptimizationPassState> per_pass_state;
  per_pass_state.reserve(registry_->passes().size());

  int num_passes_enabled = 0, num_passes_disabled = 0,
      num_passes_fallback_enabled = 0;
  for (const auto& pass_registration : registry_->passes()) {

    // Compress the operands, region arguments, and outputs.
    SmallVector<Value, 4> new_while_operands;
    SmallVector<Type, 4> new_result_types;
    new_while_operands.reserve(new_num_operands);
    new_result_types.reserve(new_num_operands);

    // Build new operands and result type.
    for (int op_idx : llvm::seq<int>(0, old_num_operands)) {
      if (removed_operand.test(op_idx)) continue;

)

func netpollinit() {
	// Unlike poll(2), WASI's poll_oneoff doesn't accept a timeout directly. To
	// prevent it from blocking indefinitely, a clock subscription with a
	// timeout field needs to be submitted. Reserve a slot here for the clock
	// subscription, and set fields that won't change between poll_oneoff calls.

	subs = make([]subscription, 1, 128)
	evts = make([]event, 0, 128)
	pds = make([]*pollDesc, 0, 128)

		return ScheduleResult{nominatingInfo: clearNominatedNode}, assumedPodInfo, framework.AsStatus(err)
	}

	// Run the Reserve method of reserve plugins.
	if sts := fwk.RunReservePluginsReserve(ctx, state, assumedPod, scheduleResult.SuggestedHost); !sts.IsSuccess() {
		// trigger un-reserve to clean up state associated with the reserved Pod
		fwk.RunReservePluginsUnreserve(ctx, state, assumedPod, scheduleResult.SuggestedHost)

}

std::vector<Output> ToOutputVector(
    const std::vector<OutputTensor>& output_tensors) {
  const int n = output_tensors.size();
  std::vector<Output> result;
  result.reserve(n);
  for (int i = 0; i < n; ++i) result.push_back(ToOutput(output_tensors[i]));
  return result;
}

// The backprop loop counter and main backprop loop run in their own execution

                                                   &hardware_map);

  // Populate the runtime metadata.
  std::vector<flatbuffers::Offset<SubgraphMetadata>> subgraphs_metadata;
  subgraphs_metadata.reserve(funcs.size());
  for (auto& func : funcs) {
    subgraphs_metadata.push_back(
        CreateSubgraphMetadata(hardware_map, &func.getBody(), &fb_builder));
  }
  auto runtime_metadata =

      << "Wrong number of arguments for function " << function_name.str();

  // Prepare function arguments from ready Chain and input Tensors.
  llvm::SmallVector<tfrt::AsyncValue*> exec_arguments;
  exec_arguments.reserve(compute->num_arguments());
  exec_arguments.push_back(tfrt::GetReadyChain().release());
  for (const Tensor& input_tensor : arguments) {
    auto av = MakeAvailableAsyncValueRef<FallbackTensor>(input_tensor);

		"overlap with any IP ranges assigned to nodes or pods. Max of two dual-stack CIDRs is allowed.")

	fs.Var(&s.ServiceNodePortRange, "service-node-port-range", ""+
		"A port range to reserve for services with NodePort visibility.  This must not overlap with the ephemeral port range on nodes.  "+
		"Example: '30000-32767'. Inclusive at both ends of the range.")

	// Kubelet related flags: