Open MihailMihov opened 1 month ago
clang-tidy review says "All clean, LGTM! :+1:"
but ReverseMode/NonDifferentiable.C is currently failing because of an issue with differentiating operator overloads in reverse mode.
If we mark that operator overload as non-differentiable, then should the issue still happen?
If we mark that operator overload as non-differentiable, then should the issue still happen?
I just tried changing the test and it still errors, but this time it is too many arguments to function call, expected 1, have 2
, while if it tries to differentiate the operator it is too few arguments to function call, expected 4, have 3
. I did open an issue for this and it is #917, where I put a more minimal failing test. I can try and look into the case where it fails when non-differentiable, but if I had to guess they are the same issue and fixing one would fix both cases.
but if I had to guess they are the same issue and fixing one would fix both cases.
Oh, yes, you are right. Thank you for the details.
fixes #717
I've added the code for handing
[[clad::non_differentiable]]
that is already present in the forward mode visitor to the one for reverse mode. I also modified the tests from forward mode, butReverseMode/NonDifferentiable.C
is currently failing because of an issue with differentiating operator overloads in reverse mode.