search

hyugit / EllipticCurve

An elliptic curve library written in Swift 4
MIT License
22 stars 5 forks source link

Why is the library not ready for production use? #6

Closed kevaundray closed 6 years ago

hyugit commented 6 years ago

hi there,

thanks for your interest. this library is very slow, because everything is written in Swift. even generating a public key takes about 30 seconds. :/

kevaundray commented 6 years ago

I really liked this library as well. 30 seconds seems like something is leaking memory or not optimised. iOS has secure enclave which uses secp256r1, so it could be the implementation?

hyugit commented 6 years ago

Yea, this library is not optimized at all. It's not using iOS native secp256r1 implementation. I wrote everything in Swift.

You can check hyu2018/UInt256 for my implementation of big number multiplication algo. UInt256's performance tests take a long time to run, especially the division algo -- because everything is written in Swift, even though I've already implemented fast multiplication/division algos. If I rewrite it in C, then it could be usable.

kevaundray commented 6 years ago

I would have no idea where to start haha. I am still impressed that you managed to code the whole thing.

Good luck in your endeavours!

kevaundray commented 6 years ago

May I ask where you got the knowledge to create this library? Is there some sort of guide online?

hyugit commented 6 years ago

I highly recommend Jimmy Song's Programming Blockchain course. Andreas Antonopoulos's book is a must for sure.

I also think this library is a good starting point. 😁 I tried to structure everything properly. Reading through the code can give you a basic idea of how elliptic curve cryptography works.

kevaundray commented 6 years ago

haha I meant the Uint256 library. You believe that this is the problem with the library right?

hyugit commented 6 years ago

oh yes. I believe the fast algos in that library can be found in major computer arithmetic textbooks. I only wrote basic ones, maybe there are better ones that I'm not aware of... but I believe the major improvement will be in rewriting them in C. (which is counter-productive, since we already have libsecp256k1)

hyugit commented 6 years ago

here is a reference book: https://members.loria.fr/PZimmermann/mca/mca-cup-0.5.9.pdf

kevaundray commented 6 years ago

Thanks for the link, have you done any tests using apple's instruments as to what is wrong with the code?

hyugit commented 6 years ago

from preliminary results, I think it's just that the 256bit multiplications in Swift are slow, because everything else is based on it. You can dive in to the stack and see for the details. If there is actually something wrong with my implementation, I'm happy to see pull requests that'll address it and solve the usability issue.

kevaundray commented 6 years ago

I would need to read up on quite a few of these topics. I will see what I can do, thanks for the links and information.