Created by: xiebaiyuan

https://github.com/PaddlePaddle/Paddle/blob/c108376506faa8c51f489a4c1e658a446424453a/paddle/fluid/operators/prelu_op.h#L44

this expression that on L44 of prelu_op.h : (temp = numel / (dim[0] * dim[1]);）
can be moved out of for loop to improve performance