#2477 Poly1305 now absorbs the whole blocks of an update straight fro…

[winne42]

…m the input with the accumulator in local variables, 1.3..1.5x as fast for 1 KB and longer

See #2477 for detailed explanation and benchmarks.

AI Disclosure

Parts of this contribution were produced with the assistance of a generative AI tool (Claude Code),
under the direction and review of the submitter, in line with the contributing guidelines. The submitter
has reviewed and understands the code. To the best of his knowledge, the contribution does not reproduce third-party material.

[dghgit]

The patch looks good, but there doesn't appear to be an AI declaration on this - it's clearly done with AI assistance though. Please see CONTRIBUTING.md

[winne42]

@dghgit sorry about that, AI disclosure note added to both ticket and PR (should it be only on one and not the other?). I found this wording of disclosure in another PR, hope this format is OK.
Maybe you could offer an issue template (and a PR template?) that already includes a formulation that you can accept?

[dghgit]

Provided disclosure is fine. Just to be clear, in our case it's not so much that we're worried about people using AI based tools submitting things, it's more that we want people to be clear that:

• they understand the need to check the code the tools are producing
• they understand that they are responsible for the code the tools are producing

AIs (at least at the moment) live in a comfortable world that is litigation proof, the same can't be said for us humans. So if a contribution results in "an issue" it's not going to be the AI the aggrieved parties will be going after, they'll base it on the code base and the commit history, so it will be us, and whoever submitted the PR.

As much as we say "misery loves company", I think everyone would agree that it's one form of company that is best avoided altogether!