variable_ptrs.reserve(num_vars);
  for (int i = 0; i < num_vars; ++i) {
    variable_ptrs.push_back(&(vars[i]->var_info));
  }
  absl::Status cc_status = LockVariables(absl::MakeSpan(variable_ptrs));
  status->status = cc_status;
}

void TF_AllocateTempForVariableInfo(TF_OpKernelContext* ctx,
                                    TF_VariableInfo* var_info,

// type, and returns an expression that yields the *runtime._type
// value representing that channel type.
func MakeChanRType(pos src.XPos, n *ir.MakeExpr) ir.Node {
	assertOp(n, ir.OMAKECHAN)
	if hasRType(n, n.RType, "RType") {
		return n.RType
	}
	return chanRType(pos, n.Type())
}

// MakeMapRType asserts that n is a "make" operation for a map type,

  TF_ASSIGN_OR_RETURN(
      auto module,
      ImportSavedModel(
          model_flags.saved_model_dir(), model_flags.saved_model_version(),
          tags, absl::MakeSpan(custom_opdefs), exported_names, specs,
          !toco_flags.enable_tflite_resource_variables(), &context, &bundle));

  if (!model_flags.input_arrays().empty() ||
      !model_flags.output_arrays().empty()) {

  tensorflow::GraphImportConfig specs;
  specs.upgrade_legacy = true;

  return ImportSavedModel(input_path, /*saved_model_version=*/1, tags,
                          absl::MakeSpan(custom_opdefs), exported_names, specs,
                          /*enable_variable_lifting=*/true, context,
                          saved_model_bundle);
}

tensorflow::Status ExportModule(mlir::ModuleOp module,

      }
      Status s;
      s = vspace.CallBackwardFunction(trace.op_type, trace.backward_function,
                                      unneeded_gradients, out_gradients,
                                      absl::MakeSpan(in_gradients));
      if (!persistent_) {
        trace.backward_function_deleter(trace.backward_function);
      }
      if (!s.ok()) {
        return s;
      }
    } else {
      if (!persistent_) {

                                                  inputs, resource_arg_indices,
                                                  &variable_infos));
    TF_RETURN_IF_ERROR(LockVariables(absl::MakeSpan(variable_infos)));
  }

  absl::StatusOr<std::vector<XlaCompiler::Argument>> args;

  if (compiler_arg_source == CompilerArgSource::TENSOR_SPEC) {

  absl::StatusOr<mlir::OwningOpRef<mlir::ModuleOp>> module =
      SavedModelSignatureDefsToMlirImport(saved_model_path, tags,
                                          absl::MakeSpan(exported_names),
                                          context, import_options, &bundle);
  if (!module.status().ok()) {
    return absl::InternalError(absl::StrCat("Failed to import SavedModel: ",

		{"Int", Const, 0},
		{"Int16", Const, 0},
		{"Int32", Const, 0},
		{"Int64", Const, 0},
		{"Int8", Const, 0},
		{"Interface", Const, 0},
		{"Invalid", Const, 0},
		{"Kind", Type, 0},
		{"MakeChan", Func, 0},
		{"MakeFunc", Func, 1},
		{"MakeMap", Func, 0},
		{"MakeMapWithSize", Func, 9},
		{"MakeSlice", Func, 0},
		{"Map", Const, 0},
		{"MapIter", Type, 12},
		{"MapOf", Func, 1},

    TF_RETURN_IF_ERROR(function_op->SetDeviceName(resource.device.c_str()));

    AbstractTensorHandle* resource_handle = nullptr;
    int num_retvals = 1;
    TF_RETURN_IF_ERROR(function_op->Execute(
        absl::MakeSpan(&resource_handle, num_retvals), &num_retvals));
    AbstractTensorHandlePtr owned_resource_handle(resource_handle);
    if (!tensorflow::isa<ImmediateExecutionTensorHandle>(
            owned_resource_handle.get())) {

		// TODO(mdempsky): Print Init and/or Body?
		if len(n.ReturnVars) == 1 {
			fmt.Fprintf(s, "%v", n.ReturnVars[0])
			return
		}
		fmt.Fprintf(s, "(.%v)", n.ReturnVars)

	case OMAKEMAP, OMAKECHAN, OMAKESLICE:
		n := n.(*MakeExpr)
		if n.Cap != nil {
			fmt.Fprintf(s, "make(%v, %v, %v)", n.Type(), n.Len, n.Cap)
			return
		}
		if n.Len != nil && (n.Op() == OMAKESLICE || !n.Len.Type().IsUntyped()) {