David Widmann

Alternatively, you could replace UnPack with SimpleUnPack (that was done e.g. in OrdinaryDiffEq) and only change the Project.toml + import statements. On Julia >= 1.7, `SimpleUnPack.@unpack` is expanded to the...

I haven't benchmarked it but that's the main idea - it does just rewrite everything using the base syntax (if available) or `getproperty`/`setproperty!` calls, so there is no difference in...

Can you include the full stacktrace in the issue?

As I mentioned in #547, `sqrt` behaves as expected if one enables NaN-safe mode: ```julia julia> sqrt(x[1]^2 + x[2]^2) Dual{Nothing}(0.0,0.0,0.0) julia> (x[1]^2 + x[2]^2)^0.5 Dual{Nothing}(0.0,0.0,0.0) ``` Honestly, I still don't...

Yes, on the master branch seeding is (again) performed without broadcasting. Depending on the structural array type the set of indices are not readily available in an allocation-free broadcastable form...

Not in its current form - it seems the PR also has merge conflicts.

This might be a duplicate of https://github.com/JuliaDiff/ForwardDiff.jl/issues/756 and be fixed by #757. Edit: Indeed, I can reproduce the problem on the master branch but with #757 I get ```julia julia>...

The PR fixes #780 as well: https://github.com/JuliaDiff/ForwardDiff.jl/issues/780#issuecomment-3373892904

Unfortunate but it is challenging to not accidentally break GPU support as long as it is completely untested - currently the package does not run a single test on a...

IIRC it was considered an internal implementation detail initially. Not completely sure but I think it was introduced to avoid depending on DataFrames.