RFR: 8349721: Add aarch64 intrinsics for ML-KEM [v7]

[Andrew Dinn]

On Thu, 10 Apr 2025 13:19:05 GMT, Ferenc Rakoczi <duke at openjdk.org> wrote:

>> By using the aarch64 vector registers the speed of the computation of the ML-KEM algorithms (key generation, encapsulation, decapsulation) can be approximately doubled.
>
> Ferenc Rakoczi has updated the pull request incrementally with two additional commits since the last revision:
> 
>  - Code rearrange, some renaming, fixing comments
>  - Changes suggested by Andrew Dinn.

@ferakocz Hi Ferenc. Thank you for adjusting the code as requested and even more so for the extra clean-ups you added which I very much appreciate.

I have added suggestions for some extra/modified commenting to clarify certain details of what is being generated that were not 100% clear to me when I first read/restructured the code. They may seem a bit obvious but I want to ensure that any maintainer who needs to review the code can assimilate it quickly (including me if/when I revisit it in 12 months time).

Mostly my recommendations for upgrading of comments is complete and I believe little more will be needed to sign off this PR. However, I still want to check through a few parts of the code that I have not fully cross-checked against the Java routines (e.g. the Barrett reductions). I'll try to do that asap but it will probably be a few days from now.

Thanks again for your help in improving this code.

-------------

PR Comment: https://git.openjdk.org/jdk/pull/23663#issuecomment-2794514677