tensorflow c++ api memory leak

When i use tensorflow c++ api in my project, GraphDef memory does not release after session->close(). i find same questions in https://github.com/tensorflow/tensorflow/issues/5302 , but use LD_PRELOAD to load tcmalloc can't solve my problem.
ENV Info
CentOS 7
CPU only
tensorflow 1.0 and python 2.7.6
Short code

for (int i = 1; i < 10; i++)
{
    GraphDef graph_def;
    Status status = ReadBinaryProto(Env::Default(), "models/graph.pb", &graph_def);
    if (!status.ok()) {
        std::cout << status.ToString() << "\n";
        return 1;
    }
    // Initialize a tensorflow session
    Session* session;
    status = NewSession(SessionOptions(), &session);
    if (!status.ok()) {
        std::cout << status.ToString() << "\n";
        return 1;
    }
    cout << "____Load graph success____!" << "\n";
    // Add the graph to the session
    status = session->Create(graph_def);
    if (!status.ok()) {
        std::cout << status.ToString() << "\n";
        return 1;
    }
    sleep(5);
    session->Close();
    cout << "----Close graph success----!" << "\n";
    sleep(5);
}

Running it and memory usage would increase。（proportional to graph file size）