Closed vhnatyk closed 5 years ago
holy shit - is it possible that moving into pure rust secp made a huge leap in performance ? In that case I will open an issue on curv to do this replacement.
Can you run test for signing and not just keygen ? in gg18 signing is much more interesting (and complex)
I wouldn't say huge - though seems a bit significant, closing this pr in favor of #9 that is just continuation on a renamed branch
Hi - restored benches from orig repo with minor refactoring (removed duplicated code in gg18 keygen.rs bench) wanted to merge it first but noticed you merged already #6 . Anyway seems from point of performance libsecp256k1 crate is ok anyway - here are some pics with times
your original multi-party-ecdsa by KZen
64 bit: x86_64-unknown-linux-gnu on nightly-x86_64-unknown-linux-gnu
prior secp256k1 crate (C based)
32bit: i686-pc-windows-msvc on nightly-i686-pc-windows-msvc
64 bit: x86_64-pc-windows-msvc on nightly-x86_64-pc-windows-msvc
Ubuntu x64 (via VM Ware Player)
64 bit: x86_64-unknown-linux-gnu on nightly-x86_64-unknown-linux-gnu
Now current libsecp256k1 crate (pure Rust)
32bit: i686-pc-windows-msvc on nightly-i686-pc-windows-msvc
64 bit: x86_64-pc-windows-msvc on nightly-x86_64-pc-windows-msvc
Ubuntu x64 (via VM Ware Player)
64 bit: x86_64-unknown-linux-gnu on nightly-x86_64-unknown-linux-gnu