cudnn Rnn : about cudnnRNNInputMode_t?

for CUDNN_LSTM :
it = sigmoid(Wi * Xt + Ri * h[t-1] + bwi + bri)
ft = sigmoid(Wf * Xt + Ri * h[t-1] + bwf + brf)

if cudnnRNNInputMode_t = CUDNN_SKIP_INPUT(make sue input_size = hidden_size), no opration is perormed at the input ,
it means like this :
it = sigmoid(Xt + Ri * h[t-1] + bwi + bri)
ft = sigmoid(Xt + Ri * h[t-1] + bwf + brf)

but I got paramer size = 8 * input_size * hiddent_size + 8 * hidden_size, not expect
4 * input_size * hiddent_size + 8 * hidden_size?

May I make a mistake for understading parameter cudnnRNNInputMode_t? who can explain this,thank you!

Yes, it is not really obvious what is meant by CUDNN_SKIP_INPUT and CUDNN_LINEAR_INPUT. I would be grateful for any explanation