Allan Leal

Hi @ibell , > Any solution that doesn't involve changes to the guts of Rosenbrock? This is a simple example, real problems involve much more complicated interplay between functions. It's...

Yes! I was too greed on perfect forwarding and this is now hurting with this use case of @ibell ! :) We might need to check the type of `L`...

If temporary `dual` (as in `dual(10) * dual(20)`) then `L&&` and `R&&` will evaluate to `dual&&`; in this case we want to make sure the multiplication results in `BinaryExpr`, and...

As discussed in the PR #68 , I agree with @ibell that that fix only partially solves this issue. However, this is mainly due to C++ inability to have a...

Hi Ian, before I answer your question, let me give you some background. `autodiff::real` is designed to be efficient in computing single-order and higher-order derivatives **along a specified direction**; it...

> Conceptually, do you think that there would be a way to make the matrix evaluation I describe more efficient with the new datatype you propose? I think it is...

Indeed, there is no way around this at the moment. Only a dedicated number type with its own specialized algorithm will be able to compute this efficiently. It would be...

I agree with you. I don't have plans to develop this capability in the near future, unless I come across the need to use this feature (maybe, in [Reaktoro](https://reaktoro.org)) or...

Hi Nukleon84, this is similar to issue #121 which seems to be false positive errors reported by vscode Intellisense. As you see, compilation works fine and you can run the...

Another thing: you may want to change `/usr/bin/g++` to a g++ with higher version (maybe you are compiling with a newer compiler (which works), but an old compiler is used...