v0.5 release goals about winterfell

[maxgillett]

It could be nice to also lift the restriction on the number of auxiliary segments in this release. The required changes are in these commits: maxgillett@4d5490f, maxgillett@f48ac44, and maxgillett@463ee45, but a test case for the library is still missing.

For the third point do you mean custom divisors? I've implemented constraint-specific transition exemptions in maxgillett@6169c6d, but it might be better for performance to make this segment-specific if there isn't a need to specify exemptions at a more granular level. There is some discussion of it in #91 (comment).

[irakliyk]

It could be nice to also lift the restriction on the number of auxiliary segments in this release.

Added!

For the third point do you mean custom divisors?

I actually meant something different: right now constraint evaluation is shared between prover and verifier (the logic is provide via evaluate_transition() functions etc. from the Air trait). This is nice because we have to write constraint evaluation logic only once and both the prover and the verifier can use it. But it also limits optimization opportunities in the prover where we know we are working with base field elements and could apply various optimizations based on that.

So, I was thinking of adding something like evaluate_constraints() function to the Prover trait. The default implementation of this function would cover these lines but then the implementers would be free to override this implementation with something different.

[cryptopunkchina]

Hi， I wonder how the the ARK1 and ARK2 generated for rp64-256 come from? I run the algorithm 5 from https://eprint.iacr.org/2020/1143.pdf can not generate the same ARK1 and ARK2

[matthiasgoergens]

When is distributed proving coming?

[irakliyk]

Superseded by #154