I observed a very small savings from increasing the optimizer runs,
and no noticable impact on compile time