AU

Thanks @smurfix !

Thanks @aavogt and @Wren6991, LGTM.

Nothing too specific. My use case is estimating probability distributions and I'd be already quite happy with jacobians and hessians of the functions from math and being able to differentiate...

> * What is a small example of a bit of code that you'd like to be able to write? Maybe something like this: ```python @jit def f(x,a,b): return exp(-a*x**2)*erf(b*x)...

@gmarkall coming back to the original question, is the description good enough? Do you need more details? @ludgerpaehler can you share the code you are working on? Do you know...

I see that there was a relevant talk in the Enzyme Conference 23 _Numba-Enzyme: A Differentiable JIT-ed Python_ Anyone here is/knows the authors and can share slides/video?

@ludgerpaehler is numba-enzyme open sourced / do you plan to open source it?

> Otherwise, what is the benefit of implementing autodiff with numba? Apart from being an interesting academic problem, of course. Uhm, weird question. The benefit is that users of numba...

One clarification: this feature request has actually three implicit parts: (1) implement something like `@CFUNCTYPE` that will map to `extern "C" { __declspec(dllexport) ...` in the generated code (2) handle...

> > (3) enable conversion of ctypes types to native pyccel types, e.g. `arr : 'int[:]' = array_from_pointer(ptr: POINTER(c_int), n: c_int)` > > I don't think (3) is possible. E.g....