1) A.mat contains a double precision variable, but in your argument list you have X declared as a float (single precision). Change that float to double.
2) A.mat contains a 20x4 matrix. The memory layout of this matrices is column-ordered. I.e., the 1st column is in memory first, followed in memory by the 2nd column, followed in memory by the 3rd column, etc. But C 2D matrices are assumed to be in memory in row order. I.e., a [20][4] matrix in C would have the 1st row in memory first, followed in memory by the 2nd row, followed in memory by the 3rd row, etc. This is not going to match up with the MATLAB memory order. If you compare your printed output you will find that, e.g., A(2,1) does not match up with X[1][0] as you might have expected. You need to think of the [ ][ ] indexing as backwards from the MATLAB indexing in order to get things to work out properly. So either you need to pass in the transpose A' like you did with I', or you need to reverse the indexing in the C routine.
