Optimizing Mobile Deep Learning on ARM GPU with TVM

Nice to have new numbers on real stuff!

great discussion about the difference between mobile and normal gpu