We have had this in our todo list for a while. The leading idea currently is to provide compiler selectable function variants that don't fall into the multiple precision abyss. The biggest challenge here is to come up with data about accuracy of results without the mp path.