Created by: qingqing01

Fix #6842 (closed)

Mainly for the broadcast in Eigen. The time changes after optimization are as follows:

• Experiments Env:

  • config: 3 stacked LSTM network, the hidden size is 64
  • 2 epoc

• Total time of 2 epoc:

  • CPU: 345.54137s -> 304.96511s .
  • GPU: 89.72162s vs 89.22058s. This optimization does not change the execution time on GPU.