}
  }

  xla::LocalClientOptions client_options;
  client_options.set_platform(platform.value());
  if (device != nullptr) {
    client_options.set_intra_op_parallelism_threads(
        device->tensorflow_cpu_worker_threads()->num_threads);
  }

  if (flr != nullptr) {
    TF_ASSIGN_OR_RETURN(auto allowed_gpus, GetAllowedGpus(flr));

  // to a simulator.

  xla::LocalClientOptions options;
  options.set_platform(platform_)
      .set_allowed_devices(allowed_devices_)
      .set_intra_op_parallelism_threads(intra_op_parallelism_threads_);
  return xla::ClientLibrary::GetOrCreateLocalClient(options);
}

Allocator* XlaDevice::GetAllocator(AllocatorAttributes attr) {
  mutex_lock lock(mu_);

  // Setting a larger thread pool does not help with the Swift caller, as we use
  // a different TFE context for each thread of execution (for running graph
  // functions, and their send/recvs corountines).
  config.set_inter_op_parallelism_threads(1);

  TF_Buffer* ret = TF_NewBuffer();
  TF_CHECK_OK(MessageToBuffer(config, ret));
  return ret;
}

TF_Buffer* TF_CreateRunOptions(unsigned char enable_full_trace) {