It produces identical/bit-equivalent results as conventional/naive matrix multiplication for integer/fixed-point data types