Fully connected layer size

Matlab trained a deep learning neural network. The structure of the net is:

1 Input Layer
2 CN Layer
3 RELU
4 Max Pooling
5 CN Layer
6 RELU
7 CN Layer
8 RELU
9 CN Layer
10 RELU
11 Max Pooling
12 FCL (size 64)
13 RELU
14 FCL (size 2)
15 Softmax
16 Classification

I saved all weight and bias matrices. The size of the input is 32x32x3 and final classes are only two. The matrix which consists the weights of the first FCN layer has size 64x4116 and i can't understand why the number of columns is 4116. The output of the second max pooling layer is 3x3x84. In my opinion the FCN layer is one row and n-columns. I set 64 neurons for the first FCN and i expext (number_of_rows)x64 however the result is 64x4116. Coud you give me some help?