There is no reason for us to enforce that things like the quadrature points/weights should be parameterized as the same type "T" that gets used for the dual. This happens in background.jl, quad_pts, quad_wts and in the ie branch where the splines are used for the iteration. Not sure if it will happen in the updated recfast.jl.
I can't imagine that doing this extra tracking is good for code performance when taking derivatives, though I am not sure how much of a difference it will make.
In any event, this is a quick fix, just putting this here so I don't forget
There is no reason for us to enforce that things like the quadrature points/weights should be parameterized as the same type "T" that gets used for the dual. This happens in background.jl, quad_pts, quad_wts and in the ie branch where the splines are used for the iteration. Not sure if it will happen in the updated recfast.jl. I can't imagine that doing this extra tracking is good for code performance when taking derivatives, though I am not sure how much of a difference it will make. In any event, this is a quick fix, just putting this here so I don't forget