}
  start_execute_.notify_one();
}

void DeviceThread::AsyncWait(TF_Status* status) {
  tensorflow::mutex_lock l(execution_mutex_);
  TFE_ExecutorWaitForAllPendingNodes(executor_.get(), status);
  TFE_ExecutorClearError(executor_.get());
}

void DeviceThread::Run() {
  while (true) {
    {
      tensorflow::mutex_lock l(execution_mutex_);
      while (execution_state_ == ExecutionState::kIdle ||

}

void TFE_ExecutorWaitForAllPendingNodes(TFE_Executor* executor,
                                        TF_Status* status) {
  status->status = executor->executor()->WaitForAllPendingNodes();
}

void TFE_ExecutorClearError(TFE_Executor* executor) {
  executor->executor()->ClearError();
}

void TFE_ContextSetExecutorForThread(TFE_Context* ctx, TFE_Executor* executor) {

//
// Note that outputs of discarded ops remain in a corrupt state and should not
// be used for future calls.
// TODO(agarwal): mark the affected handles and raise errors if they are used.
TF_CAPI_EXPORT extern void TFE_ExecutorClearError(TFE_Executor*);

// Sets a custom Executor for the current thread. All nodes created by this
// thread will be added to this Executor. It will override the current executor.

    EXPECT_NE(TF_OK, TF_GetCode(status));
    TF_SetStatus(status, TF_OK, "");
    retvals[0] = nullptr;
    TFE_Execute(matmul2, &retvals[0], &num_retvals, status);
    EXPECT_NE(TF_OK, TF_GetCode(status));
    TFE_ExecutorClearError(executor);
    TFE_ExecutorWaitForAllPendingNodes(executor, status);
    ASSERT_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);
    TFE_DeleteExecutor(executor);
  }