@Krarilotus used CPU emulation to prove that a reimplemented function behaves like the function used in the original, for example in this commit: 030b408
I personally think in cases where we have a full match, it is too much bloat, since the bytecode itself proves it.
I am undecided on whether or not to add them to the repository itself if it does not match, at least until we are in the end phase of trying to get the last functions and we really end up having to give up on some.
What I would like, however, would be a carefully written agent skill to instruct an agent to write such emulation. That may be handy for the future and some workflows.
@Krarilotus, would you like to contribute such a skill?
@Krarilotus used CPU emulation to prove that a reimplemented function behaves like the function used in the original, for example in this commit: 030b408
I personally think in cases where we have a full match, it is too much bloat, since the bytecode itself proves it.
I am undecided on whether or not to add them to the repository itself if it does not match, at least until we are in the end phase of trying to get the last functions and we really end up having to give up on some.
What I would like, however, would be a carefully written agent skill to instruct an agent to write such emulation. That may be handy for the future and some workflows.
@Krarilotus, would you like to contribute such a skill?