We consider the formulation and analysis of a regularized active-set method for large-scale convex and nonconvex quadratic programming. The method is based upon minimizing a primal-dual augmented Lagrangian function subject to upper and lower bounds on the variables. Discussion will focus on: (i) the formulation and solution of the large system of equations that defines the primal-dual search direction, (ii) the treatment of infeasible quadratic programs, and (iii) methods for estimating the optimal active set when the starting point is far from the solution. The results of extensive numerical experiments on quadratic programs from the CUTEr test collections will be presented.