PKUEcho

I just tested [GoogleNet](https://github.com/BVLC/caffe/tree/master/models/bvlc_googlenet), which should be the same model of [Tensorflow demo](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/examples/android) and it still runs slow... Did I miss anything here?

@sh1r0 Really looking forwards to your response.