Comparison of normal equation with an ABS approach for solving convex quadratic programs

In this paper, we present numerical results concerning a comparison between the normal equation approach and an ABS approach for computing the search direction of primal-dual infeasible interior point methods for solving convex quadratic programming problems (CQPs). Let m and n denote the number of constraints and the number of variables, respectively. The numerical results show that, when m/n is small, then the ABS approach needs a considerably less computing time. When m/n is close to one, then the normal equations approach is more efficient than the ABS approach.