Dev Ojha

Move multiplicative_generator from FFTField to Field

5

Currently `multiplicative_generator` is a method defined on FFTField. We should instead move this to Field, since all fields have a multiplicative generator, not just FFTField's. (Furthermore, there are contexts where...

Add AsRef<[u64]> trait bound to GroupAffine/GroupProjective ScalarField

1

This allows mul to take in an AsRef, and allows one to naturally multiply by a scalar field element. This does restrict the set of fields that can be scalar...

Optimized multiplication by constant routines

4

Something very nice that exists in Constantine is that multiplications by a constant can be inline-addition chained. https://github.com/mratsim/constantine/blob/2c5e12d5f893c6af9b97fec17d572268de3bd899/constantine/arithmetic/finite_fields.nim#L385 It would be very nice if we could do this as well....

ec: Change default scalar mul implementation to be windowed

1

Our current variable base scalar multiplication algorithm is a straight forward double and add approach. As a first order speedup, I think we should switch it to a windowed approach...

Use improved Final Exponentiation algorithms for BLS12 curves

14

Yousef recently benchmarked a variety of pairing based curve libraries' performance here: https://hackmd.io/@zkteam/eccbench, which is extremely helpful for identifying places to improve performance! Something notable is that the difference in...

Change evaluation domain API for cosets

4

Could the struct for an evaluation domain contain a Field element for the affine shift, which then enables it for use as any coset. This not existing required me to...

Fq2 neg nonresidue

1

## Description Counterpart to https://github.com/arkworks-rs/algebra/pull/207 --- Before we can merge this PR, please make sure that all the following items have been checked off. If any of the checklist items...

Investigate methods to lower the CI time

1

Currently the CI is taking on the order of 19-20 minutes to build all the crates and test them. We should see if there are thing we can do to...

I think it'd be a nice feature to provide Sage code in each curve's directory, with the fields and their generators already setup. I find myself often having to rederive...

Consider storing a pointer to the value of a coefficient in the final R1CS object

3

Loopring put out a blog post earlier today discussing optimizations they used (https://medium.com/loopring-protocol/zksnark-prover-optimizations-3e9a3e5578c0) One interesting memory reduction optimization is noting that for ~all circuits of interest, there are not that...