RFR: 8316592: RISC-V: implement poly1305 intrinsic [v7]

[Hamlin Li]

On Fri, 10 Nov 2023 18:17:09 GMT, Andrew Haley <aph at openjdk.org> wrote:

>> Ah, right, all clear! Thanks!
>> 
>> But this makes it harder or not that performant to vectorise poly3015 based on this implementation, as for example R will be R^4 or R^8 (depends on lane numbers we have at runtime), these restrictions designed by the author of poly1305 will not hold anymore.
>
> Exactly so. Vectorizing poly1305 is not easy to do at all.
> 
> The key paper is Martin Goll and Shay Gueron. _Vectorization of Poly1305 message authentication code_. In 2015 12th International Conference on Information Technology - New Generations, pages 145–150. IEEE, April
> 2015. 10.1109/ITNG.2015.28.
> 
> I have an AArch64 implementation almost ready, but it is complicated.

Thanks for sharing!
Can I ask how's the performance improvement?

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

PR Review Comment: https://git.openjdk.org/jdk/pull/16417#discussion_r1389750381