In 2008,a new Mehrotra-type predictor-corrector algorithm forlinear programming is proposed by Salahi etal.Based on cut st rategy,the algorithm implies atheoretically warranted step size in the corrector step and subsequently polynomial iteration complexity.The algorithm is ex tended to convex quadratic optimizat ion.Since the search directions aren.torthogonal any more,the complexity analysis is different from that of linear prog ramming.Using several new technical lemmas,It is proved that the algorithm,in worst case,will terminate after mathcalO(n5/2 log nE) iterations at most.In the end,a numerical experiment shows the feasibility and eff-iciency of the new algorithm.