forum.alglib.net

I have a problem that can be specified loosely as a QP problem with Box constraints, with a dimension of around 1000, but can be specified exactly as a QP problem with linear inequality constraints, with a dimension of around 10000, with around 10000 linear inequality constraints. I see from the documentation that anything over 50 linear constraints seems to be discouraged.
I have set up some code to solve the box constrained version, is it pointless to even attempt the linear constraint version? If it helps, both versions of the problem are semi-definite/convex. Any advice/gotchas I should know about?

Hi!

I think that is is better to convert problem to box constrained one. 10000 constraints is a hard problem for dense QP solver, and presently we have no large-scale sparse QP solver.