Support RISC-V

[come-maiz]

Please add support for an efficient implementation on RISC-V :D

I have been learning about the crypto and vector RISC-V extensions and they look very nice.
https://fprox.substack.com/p/risc-v-vector-cryptography-extensions

I found this implementation:
https://kernel.googlesource.com/pub/scm/linux/kernel/git/wireless/wireless-testing/+/refs/tags/v6.11-rc1/arch/riscv/crypto/sha256-riscv64-zvknha_or_zvknhb-zvkb.S

Can it be used here?

I don't think there's a board available with these extensions yet. For now I can test and benchmark in qemu and FPGA.

What this brings now, in my opinion, is a path towards decentralization in which local collectives, geographically grounded, can become the resilient foundation for the network with very efficient and transparent machines. We can start designing those machines.

[come-maiz]

In klauspost/cpuid#158 I'm exploring how to get cpuid to report these extensions.

[potuz]

This is in my backlog for sure. I am currently in the US, could pick a board but can't find one with the right extensions. A sha implementation doesn't cut it, it needs to be vectorized and streamlined for parallel hashing on trees. I'll need to study the timings and locking mechanisms of these boards and this will take me a while to do correctly and efficiently.

[come-maiz]

The [https://milkv.io/jupiter](milk-v jupiter) has vector extensions, not the crypto vector extensions.

In the Marian paper they used the AMD Virtex™ UltraScale+™ FPGA VCU118 Evaluation Kit. Very expensive. They say they are producing an ASIC, so I think it's likely that at the end of this year new boards will appear with the Zvk Vector Cryptography Extensions.

[potuz]

Fwiw, any new assembly pipelining will most likely be added to hashtree and not here in gohashtree

[come-maiz]

I see you opened the issue there. Thank you! We can talk there if you prefer to close this one.