Thank you for the feedback. Could you give us more information for investigation? When you say R outperforms SuanShu, could you send us the code so that we can do the comparison on our side? I think the speed depends a lot on the parameters you choose.
The epsilons control when the algorithm or the iterations end. One controls the precision of the embedded QP solver; the other controls the overall SQP solver. They are user specific precision parameters. There is no “mathematical” way to optimally choose them. A lot depends on trial-and-error in practice.
If you send us the code, we can see why it throws an exception. I suspect that it has to do with the epsilons you choose.