Left and right matrix multiplication using a systolic array

A method, machine-readable medium, and systolic array for left matrix multiplication of a first matrix and a second matrix are described. The first matrix is a triangular matrix, and a cross-diagonal transpose of the first matrix is loaded into a triangular array of cells in an integrated circuit. A cross-diagonal transpose of the second matrix is input into the triangular array of cells for multiplication with the cross-diagonal transpose of the first matrix to produce an interim result. The interim result is cross-diagonally transposed to provide a left matrix multiplication result, which is stored or otherwise output.