I was solving a dense and sparse variant of a QP for linear MPC. As indicated in the title I am trying to solve a dense QPs that have decision variables and constraints in the ranges specified. The operator splitting method was very fast when I tested for the sparse version of the QP but became slow when the dense QP was solved (large number of iterations to converge). I was using qpOASES till now for MPC application but in comparison to OSQP, it runs faster by an order of magnitude or higher though the solutions are less accurate than the active set solver.
Could there be any reason for this behavior? Is there any means to reduce the run time of OSQP solver for the dense QP? I am new to operator splitting methods and I would like to know what can be tuned in the solver to achieve better performance.