https://github.com/arkworks-rs/algebra/tree/master/curves/ed25519

We should've a curve25519-dalek wrapper though, because curve25519-dalek is much faster than arkworks. Arkworks exploits multi-threading, which curve25519-dalek lacks, but if you system is pegged anyways then curve25519-dalek remains a better choice.

We do have ed25519 here, but the problem here is the lack of a curve25519-dalek wrapper.

A zk prover team might only care aobut their benchmarks on dedicated prover nodes. In the polkadot ecosystem, we do care about total CPU time on validators, also now collators after elastic scaling, so the multi-threaded impl here should be worse than cruve25519-dalek.

https://github.com/nikkolasg/ark-blst looks relevant