Closed vaithak closed 1 week ago
Attention: Patch coverage is 97.53086%
with 2 lines
in your changes missing coverage. Please review.
Project coverage is 93.83%. Comparing base (
e37264d
) to head (48695ac
).
clang-tidy review says "All clean, LGTM! :+1:"
clang-tidy review says "All clean, LGTM! :+1:"
clang-tidy review says "All clean, LGTM! :+1:"
clang-tidy review says "All clean, LGTM! :+1:"
clang-tidy review says "All clean, LGTM! :+1:"
clang-tidy review says "All clean, LGTM! :+1:"
clang-tidy review says "All clean, LGTM! :+1:"
clang-tidy review says "All clean, LGTM! :+1:"
I think this is heading in a good direction. I am a bit worried that we switched from running forward and reverse to running forward twice for the diagonal only. I suspect the reason is our plan to try vector mode, but is there something else? Is there any point in making how we create the hessian configurable?
The main issue is that reverse mode currently generates a void function, which cannot be easily differentiated by any of the modes. Thus, it is beneficial to start with a forward mode for now. The second one can be anything. Similar will be true for vector mode too, and that will be a challenge to fix / make it work. So, currently, we should refrain from providing the sequence of modes as configurable.
clang-tidy review says "All clean, LGTM! :+1:"
clang-tidy review says "All clean, LGTM! :+1:"
clang-tidy review says "All clean, LGTM! :+1:"
clang-tidy review says "All clean, LGTM! :+1:"
fixes #509