How to input a matrix in an format that can be computed by a systolic array in Scala?

As we know for the two-dimensional systolic array, except that the data of the first row is sent at the time of initialization, or say the first cycle, the following ones must be sent after a row/column delay of one beat. I've tried some code for a systolic array, but it can only handle square matrices with the same rows and columns.

 for (clk <- 0 until 2*n) {
        for (id <- 0 until n) {
          val p = clk - id
          if (p >= 0 && p < n) {
            dut.io.h_in(id).poke(a(id)(p))
            dut.io.v_in(id).poke(b(p)(id))
          } else {
            dut.io.h_in(id).poke(0)
            dut.io.v_in(id).poke(0)
          }
        }
}

a,b mentioned above are two n*n matrices filled with random value. My goal is to make the values of these matrices to be assigned to h_in and v_in ports of module dut in a way that a 2D systolic array can read. So far the code works correctly.

Because of the symmetry of square matrices, the code above works, but I can't think of a feasible way to control the flow of data in a way that 2D systolic array can read when the matrix is no longer a square, such as 43, 53 or so. Can anyone give any possible guidance? Thanks in advance.