mattsignorelli

Specifically, there is no speedup nor reduction in the memory allocation for evaluation of say ```julia t = @rewrite sin(x*sin(y)) + sin(z) ```

> Correct the `@rewrite` macro works only for a limited subset of the standard arithmetic operations. Would it be a lot of work to generalize this macro for any overloaded...

Thanks. Worst case I suppose I could import and modify it. The use case is for [GTPSA.jl](https://github.com/bmad-sim/GTPSA.jl), a package wrapping a C library for manipulating truncated power series. Each mutable...

Sorry for the delay on this, I've had my hands tied with all different projects. Two weeks ago I pushed major updates/fixes and finishing touches to `GTPSA.jl`, which is now...

OK everything looks good. I had to release a new GTPSA (v1.1) to allow its jacobian to accept `AbstractArray`, and also define the `broadcastable` trait. But `Back/GTPSA` now passes all...

Sorry accidentally hit the close button instead of submit... time for bed

I submitted a PR to ADTypes here https://github.com/SciML/ADTypes.jl/pull/78

Thanks for looking this over! For second order, because GTPSA does not allow nesting, separate operators must be written. For the first order operators `gradient`, `jacobian`, there is a large...

Thanks for your work on this! > ```julia > function initialize!(xt::TPS, x::Number, dx::Number) > xt[0] = x > xt[1] = dx > return xt > end > ``` I think...

I think I understand what `Tangents` are now. Just to check: to take the pushforward of a 3 variable function with seed `[dx1, dx2, dx3]`, your `Tangents` object will be...