> Using RVV to accelerate multiple Keccak evaluations in parallel seems more promising. We will explore this and revisit the vectorization of the Keccak permutation function in a future to try to improve our current results.
I have not looked at KangarooTwelve yet, but it looks like something I should check out (https://keccak.team/kangarootwelve.html). I need to dig into BLAKE3 a bit more and I will be happy to compare notes.
> Using RVV to accelerate multiple Keccak evaluations in parallel seems more promising. We will explore this and revisit the vectorization of the Keccak permutation function in a future to try to improve our current results.
Are you going to do KangarooTwelve?! :)
I've been experimenting with a Zvbb implementation of BLAKE3, and I'd love to compare notes sometime. No doubt I'm making some silly mistakes. https://github.com/BLAKE3-team/BLAKE3/blob/guts_api/rust/guts/src/riscv_rva23u64.S
I have not looked at KangarooTwelve yet, but it looks like something I should check out (https://keccak.team/kangarootwelve.html). I need to dig into BLAKE3 a bit more and I will be happy to compare notes.