Originally posted by: Andres Codas

Hi,

I'm solving a sequence of QP problems using the concurrent optimizer.  At each iteration I remove some constraints and add others.  All Qp's are feasible by construction.

My problem is that Dual-simplex returns first, giving a solution status "unbounded", while  If I solve the problem using the barrier optimizer it returns an optimal solution.

Questions:

1. A convex QP can never be unbounded.  I guess Cplex should rather report the problem as infeasible.  Isn't it?

2. How can I make cplex wait for the solution given by the barrier optimizer in this case?.

3. If I know that the problem is bounded and feasible, and cplex is returning an inconsistent solution due to numerical difficulties, is there any way how can I control Cplex to avoid these inconsistent answers? 

Thanks!

Andres